THESE 'fi /jfr9fO.Aaf.ftOl i i j.

présentée par
Hervé GERARD
pour obtenir le grade de

DOCTEUR DE L'UNIVERSITE JOSEPH FOURIER

GRENOBLE I
(ARRETES MINISTERIELS DU 5 JUILLET 1984 ET DU 30 MARS 1992)

Spécialité: Physique

Simulation de spectres de résonance magnétique

nucléaire de cristaux liquides, polymères cristaux
liquides et polymères conventionnels

DAlE DE SOUTENANCE: 20 Octobre 1993

COMPOSITION DU JURY:

Présidont ; M. E. BELORIZKY
Examinateurs : C. NOËL
M. F. VOUNO
M. G. WEILL

•Tfcèse préparée au sein au laboratoire de Physico^himie Moléculaire

(CEA-CENG. DSM/DRFMC/SESAM) ; \

20
 THESE
présentée par

Hervé GERARD

pour obtenir le grade de

DOCTEUR DE L'UNIVERSITE JOSEPH FOURIER

GRENOBLE I
(ARRETES MINISTERIELS DU 5 JUILLET 1984 ET DU 30 MARS 1992)

Spécialité: Physique

Simulation de spectres de résonance magnétique

nucléaire de cristaux liquides, polymères cristaux
liquides et polymères conventionnels

DATE DE SOUTENANCE: 20 Octobre 1993

COMPOSITION DU JURY:

Président: M. E. BELORIZKY
Examinateurs : C. NOËL
M. F. VOLINO
M. G. WEILL

\ Thèse préparée au sein du laboratoire de Physico-Chimie Moléculaire

(CEA- CENG, DSM/DRFMC/SESAM)

21
 3

.Aux miens,

5
t'j
" Cette matière est l'étalon d'un néant qui s ignore
A. Artaud, Fragments d'un journal d'enfer. j
; .
^i
! I
" Dons toutes les tentatives faites jusqu 'à nos jours pour démontrer que 2 + 2=4,iln'a
jamais été tenu compte de la vitesse du vent. "
R. Queneau, Quelques remarques sommaires relatives aux propriétés
» aêrodynamiques de l'addition. \
•r
^\ ' *
I " C'est remplacer un mystère par un mot, dit WoJ/. Ca fait un autre mystère, c'est tout." '
B. Vian, L'herbe rouge.
*

' '!/'

22
 5

Ce travaiC s'est effectué, après maintes modifications de. son

organigramme, au Département de "Recherche JondamentaCe sur Ca Matière
Condensée du Centre d'Etudes Nucléaires de ÇrenoBCe dirigé par M. J.
Chappert, au sein du Service d'Etudes des Systèmes et Architectures
MoCécuCaires dirigé par M. M. f inert, que je remercie pour m'avoir permis de
réaCiser ce doctorat dans Ces meilleures conditionspossièCes.

A Ferdinand VoCino, qui a dirigé ce travaiC avec Ce Brio qui Ce

caractérise, je tiens à adresser ma sincère gratitude.
Au cours de ces trois années, sans ouBRer six mois de stage, iCfut un
soutien constant, me dispensant sans compter des conseiCs, des encouragements
et un savoir dont je Cui suis indéfectiBCement reconnaissant.

"Depuis deux ans, M. "P. AfdeBert, qui a succédé à M. Tineri dans Co

comBien accaparant fauteuiC dé directeur de CaBoratoire, a su gérer (équipe de.
Physico-Chimie MoCécuCaire avec un zèCe et une pugnacité dont je Cui suis
également reconnaissant.

£n acceptant de participer à Ca Commission d'Examen en tant que,

respectivement, président et rapporteurs, M. "E. 'BeCorizky, Mme C. WoeCet M.
Q. yyeiCC méritent toute ma gratitude pour Ce temps précieux et Ca BienveiCCance
accordés aujugement de ce travaiC.

"Bien des résuCtats présentés ici, à vrai dire Ca pCupart, n'auraient pu

être oBtenus sans Mme et M. IL1B. et A. "BCumstein qui ont fourni CessentieCdes
échantiCCons poCymères, ni sans Ca coCCaBoration active et enthousiaste de MM.
"D. QaCCand, J.3. yerreira et J. AvaCos.
Vn grandmerci à eux tous.

foute ma reconnaissance est exprimée ici à M. T. fries pour son affaBCe

et permanente disponiBidté à Coccasion, entre autre, de nomèreux proBCèmes
informatiques.

"Et même, si nos coCCaBorations n'ont pas toujours produit Ces résuCtats
escomptés, je tiens à remercier pour Ceur aide scientifique et technique MM. J.
i"» Langawsèi, J. Laugier et J. MouCin ainsi que tous Ces memBres du (groupe
\ Informatique, enparticuCier J. Merdrignac et 3. Thomas.

• \ Comment, maintenant, exprimer ma gratitude envers toute [équipe de

• "Physico-Chimie MofecuCaire?
O muses de Ca physisco-chimie, je commencerai par vous, en remerciant
^ 7. AndbCfatto pour sa constante Bonne humeur, C. "Barthet pour son charme
\ franc-comtois, A.M. "Outran pour ses Bonnes manières, S. "EscriBano pour sa
r résistance aux agressions verBaCes et corporeCCes de ses compagnons de Bureau,
J M. ÇjugCieCmi pour ses campagnes de propreté, JfathaCie M. pour Marc et
* Sophie, Annette pour sa vitaRté percutante (aïe, pas Ca tête!), J.JL "RMtO pour

23
 son deCicieux accent américain et S. 'Ristori-'Pedbcchi pour Ce trouBCe qu'eCCe a
suscité dans ce CaBoratoire.
Messieurs, c'est votre tour, aCors merci à Jean-Jacques pour son
magnétisme animaCà Ca. (imite de (a BestiaCité, 1P. AudéBertpour sa phiCogyni&
persiflante, M. "Brotte pour sa Bonhommie fCeurant Ca Cavande et Canis, J.T.
CasquiCho pour sonfCegme lusitanien, 1E. Claude pour son engagement en faveur
de ta restauration du Royaume des francs, Jean-yves pour ces fCeurs, parfois
rhétoriques, qu'if fit écCore étrangement dans nos Bureaux, 3ferr Ttoktor QeBeC
pour Ca préservation des ponts de Ca Drame, A. de Çeyerpour Cécran qui m'a
permis de rédiger cette thèse, 1P. gros dMCCon pour son Bronzage, Benoît
Loppinstrômpour Ce surstrômming et sa Cutte perpétueCCe contre Mister 3oyer,
y. MaréchaCpour sa casquette, francky "Je-me-présente-francky", dit Ouin-
Ouin, pour son oeuvre, S. Miachonpour son superBe parapCuie. BicoCore et sa
cuCture encyclopédique (ruÉrique pComBerie), "Renato M. pour ClrCande, ses
Ceprecfiauns et Ce foie de voCaiCCe aux auBergines, f. "J'ai-pas-Bien-compris"
WoveC-Cattin pour nos Baignades irlandaises communes, T, "Snake" Stevens
pour Ca BCague du BeCge dans Ca patinoire, "P. Terecfipour Ces poCémiques au
sujet de Ca prononciation de son 'cK nominaCfinaC et 3f. ZeCsman pour ces
années de coCtaBorationfructueuses avec nos amis d^Outre-'Rfun.
Monument incontournaBCe s'éCevant au troisième étage dit Bâtiment Cs,
QiCBert "Baud-rand, sans CequeC, on ne Ce criera jamais assez, ce CaBoratoire ne
serapCus ce qu'iCa été, que ton Nom soit sanctifié et BriCCe en Cettres d'or au
paradis de Ca torréfaction.
"Etpuis, surtout, n'ouBCionspas Mmes M. "BurCet, y. Martin, C. "Petit et Z.
Termacfie qui, teCCes des infirmières attentives, se sont succédées au chevet de
TCM, résoCvant avec une efficacité et une Bonne voConté jamais démenties nos
tracas quotidiens.

ConcCuons en remerciant ceCCes et ceux qui m'ont supporté pendant pCus

de trois ans à ÇrenoBCe et aiCCeurs, je parCe Bien entendu de CdantaC, Coco,
Vjamila, française, Lucia et Jeff ainsi que Ces frères de Ca Côte et tous Ces 1BoB
et lîoBette désormais éparpiCCés aux quatre coins d'un hexagone amputé.

Ainsi s'achève, pour une provisoire escaCe, Ce "Voyage....

24
 AVANT-PROPOS

Ce travail repose sur l'exploitation de spectres RMN en phase

fluide anisotrope (nématique essentiellement) effectuée en décomposant
les propriétés observées (les interactions magnétiques moyennes) en
deux aspects complémentaires, l'aspect moléculaire déterminé par la
géométrie et les mouvements internes à la molécule incriminée, et
l'aspect collectif qui prend en compte tous les mouvements externes
auxquels est soumise cette molécule.
Le présent mémoire se décompose en trois parties principales.

Dans une première partie sont exposés les bases théoriques du

travail ainsi que les deux modèles utilisés pour décrire l'aspect
moléculaire (modèle à conformation unique) et l'aspect collectif
(théorie de l'influence des modes élastiques sur les formes de raias
RMN). Ce dernier modèle a fait l'objet d'une publication qui est
reproduite à la fin de cette première partie et précédée d'un résumé
des hypothèses utilisées, des problèmes rencontrés et des résultats
théoriques obtenus dans cette étude.

Dans la deuxième partie sont ensuite résumés des résultats

provenant de l'application de ces modèles à des nématiques de faible
masse moléculaire (biphényls substitués en solution dans un nématique,
pentyl-cyanobiphényl pur en phase nématique) et des polymères cristaux
liquides linéaires à chaîne principale. Ces résultats ont eux aussi
fait l'objet de publications qui sont jointes à cette deuxième partie.

Ces modèles destinés à traiter les données de RMN des nématiques

sont étendus au cas de polymères conventionnels. La description
détaillée des résultats, non encore écrite sous forme de publication,
fait l'objet de la troisième partie de ce travail.

5
I
Vi
J

25
 9

SOMMAIRE

INTRODUCTION GENERALE 11

I CONSIDERATIONS THEORIQUES 17

I A Rappels sur les phases cristal liquides et

leurs propriétés visco-élastiques 19

I A l Phases cristal liquides 19

I A 2 Propriétés visco-élastiques des nématiques 20

I B Principes généraux de la RMN des phases

anisotropes fluides 27

I B l Rappels 27

I B 2 Le tenseur d'ordre 27

I C Modèle à conformation unique 31

I C I Principe 31

I C 2 Expression des interactions magnétiques 32

I D Influence des fluctuations visco-élastiques

sur les formes de raie RMN 33

I D l Généralités 33

I D 2 Elargissement de raies RMN par les modes

élastiques 33

••t

.3-1

27
 10

ïï CRISTAUX LIQUIDES 59

n A Biphényls substitués dans des solvants

nématiques 61

U A 1 Exploitation des données RMN 61

H A 2 La méthode de maximum d'entropie 61

n B Pentyl-cyanobiphényl 71

E B l Analyse des données RMN 71

n B 2 Exploitation des résultats provenant

de la RMN 107

il C Structure et ordre orientationnel

de polymères nématiques linéaires 135

M POLYMERES CONVENTIONNELS 147

Dl A Simulation de spectres RNN d'un polymère

rigide 149

M A I Généralités 149

M A 2 Spectres alignés du polymère en

solution dans un solvant nématique 150

DI A 3 Spectres du polymère pur 153

if*
j ,t
m B Simulation de spectres RHN d'un polymère
serai-flexible 159

i CONCLUSION 165
<r
\ REFERENCES BIBLIOGRAPHIQUES 169

28
 INTRODUCTION

GENERALE

'!

29
 13

Le mariage, consommé depuis une trentaine d'années, entre la

méthode de résonance magnétique nucléaire (RMN) et la phase nématique
s'est révélé une union des plus fructueuses[1-6]. L'origine de cette
réussite matrimoniale réside dans le fait que la phase nématique se
caractérise par un ordre orientât!onnel à longue distance selon un
vecteur directeur ï£ et que la RMN est sensible aux propriétés,
interactions et mouvements des spins par rapport à un champ FC. En
champ fort, ce qui est le cas lors d'expériences RMN, le directeur n^
correspond pour la plupart des nématiques au champ extérieur statique
FC. Ainsi, pour ces nématiques, la RMN va fournir des informations sur
l'orientation et le mouvement des molécules par rapport au directeur
n^. Cependant l'exploitation de ces informations provenant de données
RMN diverses nécessite l'utilisation de modèles et reste actuellement
le théâtre de farouches controverses[6].
Jusqu'à maintenant, les modèles utilisés dans ce dcmaine
s'attachent surtout à décrire les données RMN des nématiques de faible
masse moléculaire, et la plupart de ces modèles sont
multi conformât!" onnel s [1-5]. L'idée de base, assez répandue dans le
; milieu des cristaux liquides, sous-tendant de tels modèles est de
considérer la phase liquide isotrope de la même manière que la phase
gazeuse, les molécules y possédant la même liberté conformationnelle,
et les phases liquides anisotropes comme des phases gazeuses
partiellement orientées, cette orientation partielle limitant
également les possibilités conformât!onnelles des molécules.
Cependant, une autre vision de la phase nématique, considérant
celle-ci comme une phase solide où les corrélations transiationnelles
disparaissent à longue distance, a récemment fait ses preuves [6-8].
Il s'agit du modèle à conformation unique.
Dans ce qui suit est présenté et testé un modèle permettant de
simuler les spectres RMN des nématiques, polymères ou non, mais
pouvant également s'appliquer à des phases amorphes ou solides, voire
aux liquides usuels. Le principe en est le suivant: obtenir le spectre
théorique correspondant à une (petite) molécule, ou à un oligomère
dans le cas de polymères, puis y surimposer les effets de mouvements
externes collectifs.
Ce spectre théorique prend en compte les principaux mouvements et
^ interactions magnétiques intramoléculaires, les interactions
J
j intermoléculaires étant négligées en raison de Tisotropie et la
rapidité des mouvements transiationnels, du moins en phases nématique
1 e
I t isotrope. Afin de le calculer, le modèle suppose une conformation
--' unique (en fait plusieurs conformations équivalentes, d'un point de
vue RMN, du fait de l'existence de possibles mouvements internes
correspondant à des opérations de symétrie). L'idée sous-tendant ce
\ modèle est le fait que les densités varient peu entre les phases
j) solide et nématique (et liquide), d'où un arrangement local très
1 semblable dans toutes ces phases et des possibilités
- conformât!onnelles peu éloignées. Même si l'existence de différentes

30
 14

>
t conformations est admissible, et souvent utilisée dans diverses
analyses, cette hypothèse .se révèle intéressante dans sa simplicité
même, cette conformation unique pouvant toujours être perçue comme la
"moyenne" d'une distribution de conformations. Cependant les tests
récents de ce modèle sur diverses moiécules semblent montrer q u ' i l est
plus qu'un modèle opérationnel mais correspond à une réalité physique
dans la mesure où ces tests correspondent à des situations où le
nombre de paramètres ajustables est bien moindre que le nombre de
données expérimentales indépendantes [6],
Reste alors à prendre en considération les mouvements collectifs
externes auxquels est soumise la molécule. Les mouvements de
translation, dont l'effet n'est perçu que via les interactions
intermoléculaires, n'entrent comme on Ta déjà indiqué que pas ou peu
en ligne de compte dans les phases nématique et liquide. Cependant, on
ne pourra pas rigoureusement les négliger si l'on veut traiter clés
spectres solides ou amorphes. Les mouvements rotationnels sont eux
bien décrits par la théorie des modes visco-élastiques[9], dont nous
allons user pour exprimer l'influence de ces mouvements sur la forme
de raie RHN.
Pour les cristaux liquides de basse masse moléculaire, ces
fluctuations sont en général assez rapides pour avoir sur le spectre
un effet de rétrécissement, le spectre correspondant ne différant du
spectre théorique que par un simple facteur d'échelle dans le cas de
molécule présentant une faible biaxialité. Pour les polymères cristaux
liquides (PCL), par contre, certaines fluctuations visco-élastiques
sont suffisamment lentes pour avoir un effet notable d'élargissement.
Un modèle, précédemment proposé par Esnault et al.[9], divise ces
fluctuations en deux groupes (infiniment rapides ou statiques sur
l'échelle de temps RHN de l'expérience considérée) et introduit la
notion de paramètre d'ordre statique. Un nouveau modèle, qui tient
compte de toutes les fluctuations visco-élastiques et de leur temps
caractéristique selon la théorie des modes visco-élastiques, est
développé dans cette thèse. Tous ces aspects concernant le traitement
de spectres RHN seront développés dans le premier chapitre.
Nous verrons ensuite l'application du modèle à conformation
unique dans le cas de molécules de faible masse (pour lesquels les
fluctuations visco-élastiques ne se manifestent qu'à travers le
i*» tenseur d'ordre). Ces molécules sont d'une part deux biphényls
| substitués en solution dans un solvant nématique et d'autre part le
t '. pentyl-cyanobiphényl (5CB), un cristal liquide nématique. L'analyse
;
des nombreuses données RHN du 5CB a mené à un résultat en
contradiction apparente avec le postulat de base du modèle, mais non
avec son idée de départ. De plus, cette analyse a permis
l'exploitation de données autres que celles de la RMN. Ensuite seront
V analysés, toujours dans le cadre du modèle à conformation unique, les
f spectres de RHN du proton de polymères nématiques et de leur molécule
\ Bodèle. L'influence des modes élastiques sur les formes de raie d'un
de ces polymères fait l'objet de l'application dans l'article en

3l
 15
//6
exposant la théorie et reproduit à la fin du chapitre I. Ces
applications du modèle de conformation unique à des cristaux liquides,
de faible masse moléculaire et polymères, feront l'objet du deuxième
chapitre.
Enfin le troisième chapitre sera consacré à l'extension de
l'étude aux polymères ne présentant pas de phase nématique, pour
lesquels nous supposerons que, localement, l'ordre peut être considéré
comme nématique, l'absence de corrélations orientationnelles à longue
distance revenant à considérer ces matériaux d'un point de vue
macroscopique comme un ensemble de monodomaines nématiques dont les
directeurs respectifs présentent une distribution isotrope. Pour Tun
des deux polymères traités, un polymère rigide, l'étude de la phase
pure est précédée et facilitée par l'analyse de spectres de ce
polymère en solution dans un nématique de faible masse moléculaire.
Pour l'autre polymère étudié, un polymère semi-flexible, la base de
départ de la simulation provient des données cristallographiques.

i
J
\
H
ï

.) . v
}

35
 CHAPITRE I

CONSIDERATIONS

THEORIQUES

36
 19

I A Rappels sur les phases cristal liquides et leurs propriétés

vi sco-élastioues f101

I A l Phases cristal liquides

Les cristaux liquides, découverts en 1888 par le botaniste

Reinitzer[ll], présentent des états mésomorphes intermédiaires entre
les phases solide et liquide. On distingue deux familles de telles
mésophases, les cristaux liquides lyotropiques, présentant cette phase
en solution pour une certaine gamme de concentrations, et les
thermotropiques présentant, lorsqu'ils sont purs, une phase cristal
liquide pour un domaine de température sis entre les phases solide et
liquide.
D'un point de vue structural on peut classer ces phases
mésomorphes en trois catégories: phases nématique, smectique et
cholestérique.
* La phase nématique possède un degré de liberté de moins par
rapport à la phase liquide. Les molécules y sont orientées de manière
préférentielle selon un vecteur directeur n 0 .
* Les phases smectiques présentent également une orientation
préférentielle des molécules selon un vecteur directeur mais possèdent
un degré de liberté trans.lationnelle de moins que la phase nématique.
Les centres de gravité des molécules y sont en effet arrangés selon
des plans parallèles. Dans le cas des smectiques A et B, ces plans
sont perpendiculaires à n^.
* La phase cholestérique, enfin, est assimilable à une phase
nématique dans laquelle le vecteur directeur varierait hélicoïdalement
dans l'espace selon une direction perpendiculaire à n^ (voir figure
1).

\
1
>V '
figure 1: schéma des phases a) neraatique
b) smectique
\ c) cholestérique
 20

Dans chacune de ces phases, s'il existe une orientation

moléculaire préférentielle selon n^ (ce directeur étant spatialement
variable dans les cholestériques) au niveau macroscopique, au niveau
microscopique les mo^cules fluctuent autour de ce directeur du fait
de l'agitation thermique. Afin de caractériser ce désordre, et pour
des molécules présentant une symétrie cylindrique et des mésophases
uniaxiales (telle .a phases nématique), on peut alors définir un
paramètre d'ordre S,

S = < P 2 (cose) > = T < 3cos29 - 1 > (D

où e est l'angle entre l'axe de la phase mésomorphe (le vecteur

directeur) et l'axe de la molécule et où <> indique une moyenne sur
l'ensemble des molécules. Le choix du polynôme de Legendre de deuxième
ordre se justifie par le fait qu'il remplit les conditions classiques
pour caractériser la transition mésophase-isotrope (S=O en phase
isotrope, S>0 en deçà) et il va de plus intervenir dans le cas de
propriétés physiques faisant appel à un tenseur de rang 2, telles les
interactions prises en compte en RMN.
Cependant, la plupart des molécules présentant des mésophases
uniaxiales ne possédant pas la symétrie cylindrique, Tordre devra
alors être caractérisé par un tenseur d'ordre décrivant le désordre
moyen d'un référentiel moléculaire par rapport à r£.

I A 2 Propriétés visco-ëlastiques des nématiques

I A 2 a Propriétés statiques

L'état du fluide nématique est caractérisé par deux grandeurs

associées à l'unité de volume située au point r, et ceci à un temps t
donné: la vitesse v(r,t) et le vecteur directeur local n(r,t). A
l'équilibre, on a une orientation uniforme du directeur n(r,t) = n^.
A toute distorsion du champ d'orientation va correspondre une
énergie élastique emmagasinée, l'énergie de Franck (par unité de
volume):

FdV
*. •
où V est le volume de l'échantillon.
Si l'on ne considère que de petites déformations, on a:
2F = K,(divn)2 + K2(n.™tn)2 (2)
où les K1 sont les constantes élastiques associées aux trois
déformations élémentaires qui sont, respectivement, Tébrasement
(splay), la torsion (twist) et la flexion (bend) (voir figure 2).
 •-.;,&.

21

X •

! /,
U^

(2)

Bind

(3)

NÎ
Twill

'!
Vi

figure 2 : Les trois types de déformations élastiques: 1) ébrasement

2) flexion 3) torsion.
\
 22

En présence d'un champ magnétique H, on doit ajouter à l'énergie

volumique F le terme F1n - - 1/2 xa (n-H) 2 où xa est l'anisotropie de
susceptibilité magnétique par unité de volume.

I A 2 b Propriétés dynamiques

L'anisotropie du fluide nématique impose, en addition à

l'équation de Navier-Stokes gérant les mouvements des centres de
gravité des molécules, une seconde équation régissant les mouvements
de rotation du vecteur directeur[12].
La première loi est de la forme:

(3)
dt ax.

o est le tenseur des contraintes appliquées au matériau:

t t t
pression tenseur tenseur
élastique visqueux

p est la pression, hydrostatique et provient de l'énergie

élastique de Franck:

Le tenseur s'écrit:

06 n A
j = I k "p (cp "i "j +

a4 A1J + Ot5 nd nk Akl

avec ta = - rot v
é*
'1 Les coefficients O1 (coefficients de Leslie) ont la dimension
d'une viscosité et «4 représente la viscosité ordinaire associée à un
Vl fluide isotrope. Seuls cinq de ces coefficients sont indépendants car,
selon la relation de Parodi[13], on a:

+ Ot = O6 -

Certaines combinaisons de ces coefficients peuvent être

\ directement mesurées dans les expériences de cisaillement donnant
accès aux viscosités de Hiesowicz[14].

I ':/"
 23

La deuxième équation gouverne les rotations du directeur qui

subit l'action de deux couples, un couple élastique interne T1 qui
tend à le ramener selon sa position d'équilibre et un couple visqueux
F2 traduisant le transfert du mouvement de rotation local au mouvement
hydrodynamique global. Ces deux couples s'expimenï «Je I^ manière
suivante:
d df
-> ( ]
n A Ûh ou-
— ^79x. } J
et

T2 = n A (-Y1 . iî + Tr2 A n)

avec -Y1 = Ot3 - O2 et -Y2 = O6 - as

En définissant n = n A (dn/dt) la vitesse de rotation du

directeur, on a alors:

(4)
Ht
Enfin, l'hypothèse d'incompressibilité du fluide fournit la
troisième équation:

= O (5)
SX,

I A Z c Fluctuations thermiques d'orientation

Le Groupe d'0rsay[15a] a développé un modèle permettant, à partir

des équations hydrodynamiques précédentes, d'obtenir le spectre en
fréquence des fluctuations thermiques du vecteur directeur n(r,t).
En faisant l'approximation de petites déformations ( => d/dt =
s/at = iw, 3/Sx1 = Iq1 ) et en supposant que les mouvements de n sont
lents à l'échelle moléculaire ( => dn / dt négligeable), (3), (4) et
(5) deviennent alors:

lu P v = a q

n A h = T2 \ (6)

q . v =O

En supposant maintenant que les fluctuations &n du vecteur

directeur sont telles que (approximation linéaire)

n(r) = n0 Sn et n0 . 6n = O
\
 24

Flexion

Mill
\\\\\\
Flexion

figure 3 : repère {ê7,ë^,ï£) découplant les modes propres de

. fluctuations d'orientations (haut) et (bas):
a) déformation du mode de polarisation a=l (splay-bend)
\ b) déformation du mode de polarisation o«=2 (twist-bend)
 .-.lit

25

avec
HT(O1O1I) , 3(O1O1H) soit Sn(nx,n ,0)

l'équation (2) devient, en négligeant les termes d'un ordre supérieur

à 2 en Sn:

+ Xa H2 (nx + n 2 )

En passant aux composantes de Fourier spatiales (n^(q)

_ J H1(Ir) exp(iq.r) dr) et en travaillant dans le référentiel (ê7,ë
n7 ) tel que &J* appartienne au plan (q,n^) avec ë7.n7=0 et
orthogonal à ce même plan (voir figure 3), on a alors:

2F 2
= Z S ma(q)l2 (K 3q2 + XaH ) (7)

Pour chaque vecteur d'onde q coexistent deux modes représentés

sur la figure 3, le mode a=l mêlant les déformations d'ébrasement et
de flexion (splay-bend) et a=2 mêlant les déformations de torsion et
flexion (twist-bend).
La formule (7) permet d'exprimer K et de résoudre le groupe
d'équation (6) afin d'obtenir les fréquences propres des modes de
fluctuation na(q). Pour un vecteur d'onde q et un a donnés, deux
modes coexistent, l'un rapide et l'autre lent (ceci dans la limite des
champ faibles et pour KpAi2 ^L où K et t) sont les ordres de grandeur
des K1 et oc- ). Dans ce qui suit, nous nous préoccuperons
essentiellement des modes lents dont les temps de relaxation sont
donnés par:

(8)
Mi + M2 + xaH2
où les coefficients de viscosité -na(q) sont expressibles en fonction

des coefficients Ot1 de Leslie et de q.

Enfin, à partir de (7), et en appliquant le théorème

d'équipartition de l'énergie, on obtient:
kT
(9)
<
\
 27

I B Principes généraux de la RMN des phases anlsotropes fluides

I B 1 Rappel s[16]

Pour un spin isolé en présence d'un champ magnétique statique H 0 ,

les niveaux d'énergie accessibles à ce spin sont déterminés par les
valeurs propres de de ll'hamiltonien Zeeman, de la forme E m = -mfrvH0
(m=-I,I), où -Y est le rapport gyromagnétique du spin concerné. Pour
ûm=l , la transition correspondra à une pulsation de résonance
(pulsation de Larmor) ( = -A-
Cependant, dans la matière condensée, le champ extérieur perçu
par un spin, et donc ses états d'énergie possibles, seront fortement
influencés par son environnement.
Outre le terme Zeeman, on considérera les interactions
magnétiques suivantes, de nature tensorielles, venant s'ajouter à
l'hamiltonien de spin du système:
* le déplacement chimique, qui répercute l'influence du champ
extérieur sur les spins nucléaires via leur cortège électronique. Cet
effet, résultant de l'environnement électronique, est donc une
signature propre à chaque groupement chimique auquel appartient le
spin nucléaire observé et dont l'amplitude est proportionnelle à celle
du champ H0*.
* l'interaction dipolaire, résultant du couplage dipolaire entre
deux spins nucléaires et ne dépendant pas de l'amplitude du champ
appliqué mais du module et de l'orientation par rapport au champ du
vecteur joignant les spins considérés.
* l'interaction quadrupolaire, provenant de l'interaction entre
un spin nucléaire et le gradient de champ électrique dû à son
environnement électrique. Cette interaction ne dépend pas de
l'amplitude du champ magnétique extérieur et est observée pour des
spins supérieurs à 1/2 . Pour ceux-ci, elle est alors prépondérante.
D'autres interactions existent mais dont l'intensité est
négligeable par rapport aux trois interactions déjà mentionnées, du
moins en phases nématique et a fortiori solide ou amorphe.

I B 2 Le tenseur d'ordre orientationnel

Le spectre RMN observé à haut champ dépendra de la projection

selon le champ statique H0 des divers tenseurs moléculaires du second
ordre représentant les interactions précédemment mentionnées. La
valeur de cette projection dépend de l'orientation de la molécule
vis-à-vis de ce champ. De plus, l'agitation thermique va faire varier
cette orientation durant le temps de mesure, d'où l'observation d'une
projection moyenne. Il est donc utile d'introduire ici la notion de
tenseur d'ordre orientationnel, précédemment évoquée, qui va décrire
 28

la désorientâtion moyenne d'un référentiel moléculaire par rapport à

-t
un axe n , lié au referential du laboratoire, donné (dans le cas des
nématiques, le vecteur directeur n^ ). On définit ce tenseur T de la
manière suivante:

-< >/ (10)

où e.,=(i,n) , i=x,y,z étant l'un des trois axes du référentiel

moléculaire (voir figure 4).
Ce tenseur est symétrique et de trace nulle. Par définition, et
par la suite nous noterons S et S1^ le tenseur d'ordre et ses
composantes exprimés dans son référentiel principal, et T et T1J dans
les autres cas. Par convention, on définit le repère principal (X,Y,Z)
de façon que les valeurs propres répondent à la condition
s
zz *5Xx^ SYY • L'axe Z correspond à l'axe moléculaire Témoins
-C
désorienté vis-à-vis de n, et, dans son repère principal, S est défini
par deux valeurs propres indépendantes: le paramètre d'ordre uni axial
S2 qui représente l'orientation moyenne de Z par rapport à n et
SKX " SYY Q u i traduit la biaxialité de Tordre moléculaire. La
projection moyenne Qrr d'un tenseur moléculaire Q, correspondant à des
interactions magnétiques ou à d'autres propriétés (susceptibilité
magnétique ou biréfringence optique par exemple), selon un axe n
(celui par rapport auquel on définit le tenseur T) sera alors:

Qm = < n|Q|n >

En exprimant cette équation dans le référentiel moléculaire

associé au tenseur d'ordre, on a alors:

<COS8
i COS
V (U)

Or, selon (2) on a:

<COSGI cose^ = ( 2T1^ + S1J )/3 (12)

1
J*
i d'où

soit
V-
if* = Tr(Q) + Tr(Q.T)
nn (13)
 29

Dans le cas d'un ordre orientationnel

-t- parfait en l'absence de
fluctuation, Taxe Z sera parallèle à n avec S z z = 1 (S xx = S Y Y = -1/2)
et on aura simplement Qnn = Qzz . Dans le cas opposé de larges
fluctuations rapides et isotropes (cas d'un liquide isotrope), on aura
T 1 J - S 1 1 - O B t Q11n- j Tr(Q) =0".
On constate donc que pour un monodomaine nématique où les
fluctuations autour du vecteur directeur sont rapides par rapport au
temps d'observation, la quantité observée dépend de deux facteurs: une
contribution moléculaire (le tenseur Q) et une contribution
statistique (le tenseur d'ordre T). Pour analyser des données RMN, il
sera donc utile de prédire les tenseurs Q associés aux spins concernés
grâce à un modèle approprié afin de déterminer ensuite le tenseur T à
partir de l'équation (1). Un des modèles moléculaires possibles est
présenté dans le paragraphe suivant.

'\

, figure 4 : Illustration des axes et angles utilisés pour définir

\ l'ordre orientationnel moléculaire. Le vecteur directeur
macroscopique est identifié par N.
 30

Le problème est en fait souvent plus complexe dans le cas où

l'interaction dipolaire est prépondérante (spin = 1/2). Dans ce cas,
la détermination des Q11n, traduisant les interactions magnétiques,
n'est pas directe mais doit se faire à l'aide de la simulation d'un
spectre complexe. Ce genre de calcul peut être réalisé à l'aide de
l'algorithme développé par Gall and et Ferreira[8a], le calcul étant
limité par le nombre de spins interagissant à prendre en compte. Ceci
implique, vue la limitation actuelle des possibilités informatique (on
ne peut traiter un nombre de spins supérieur à 14 sur Cray n), de
négliger les interactions intermoléculaires (ce qui est raisonnable en
raison des mouvements transiationnels dans les phases fluides) et de
ne traiter que le cas de molécules possédant un faible nombre de spins
ou des éléments de molécule, contenant des petits groupes de spins 1/2
relativement isolés (par deutération sélective, par exemple). Dans le
cas des protons il est cependant fondamental de traiter le problème
exactement (ce qui nécessite la diagonalisation de matrices 2N x 2N où
N est le nombre de spins interagissant pris en compte) car Terreur
faite en ayant recours à des approximations (considérer deux groupes
de sept spins plus les perturbations au premier ordre entre les deux
groupes au lieu d'un groupe de quatorze spins, par exemple) peut être
du même ordre de grandeur que les effets que l'on veut étudier.

«."*

•1

\
 i C Modèle à Conformation Unique

I C l Principe

Le modèle utilisé dans cette étude pour le calcul d'interactions

magnétiques et de spectres en résultant est Is modèle dit à
conformation unique[6,8]. Le principe en est simple. On y suppose que
dans la phase nématique, à une température donnée, les molécules
n'adoptent qu'une seule conformation (en fait deux, si la molécule est
chirale: la conformation "unique" et son image dans un miroir). Le
terme de conformation unique peut donc apparaître un peu abusif: les
conformations existant au sein d'une phase nématique, suivant ce
modèle, peuvent être décrites par un ensemble unique de paramètres
géométriques, et le désordre orientationnel exprimé par un unique
tenseur T. Ainsi, l'équation (13), qui s'appliquait à une unique
molécule, va être applicable à un monodomaine nématique (ou autre)
entier. D'un point de vue dynamique, tandis que les mouvements
externes collectifs seront pris en compte par le tenseur d'ordre (dans
l'hypothèse de mouvements suffisamment rapides), le modèle prendra en
compte les mouvements moléculaires internts.
Les mouvements internes considérés seront des opérations de
symétrie, laissant la conformation inchangée, telles la rotation de TT
d'un phënyl autour eb son axe para ou celle d'ur groupe méthyl autour
de son axe 3. L'autre opération de symétrie importante sera le
mouvement J'échancd entre une conformation et son image dans un miroir
(mouvement de racémisation) par rotation simultanée de blocs
moléculaires rigides autour des simples liaisons. Il est évident que,
pour que cette opération soit une opération de symétrie, les blocs
moléculaires considérés doivent être achiraux.
Il existe d'autres mouvements internes telles les fluctuations
rotationnelles autour GP* liaisons. Ces librations sont considérées
être d'assez faible amplitude pour avoir un effet de moyenne
négligeable sur les interactions magnétiques. La conformation unique
doit alors être perçue comme la moyenne sur ces librations.
f* L'idée de base sous-tendant ce modèle est l'identification de la
! phase nématique (et, par extension, de toute phase fluide) à une phase
t solide ayant perdu ses corrélations à longue distance (la phase
\ nématique conservant néanmoins de telles corrélations
orientationnelles). Cette identification est supportée par le fait que
la densité varie assez peu (quelques %) entre les états solide,
s néaatique et même liquide, ainsi que par les expériences de
» diffraction révélant des pics, certes très élargis, à des angles
f voisins des pics de Bragg du solide correspondant. La conformation
« ainsi que l'arrangement local des molécules doit donc peu différer
d'un état à l'autre, Io transition solide- isotrope, par exemple,

48
 35

l'on suppose que le volume d'untégration correspond à un cylindre

circulaire on distingue en tout quatre paramètres de coupure, q^a et
q. . Cependant le nombre de modes dans un monodomaine nématique (de
volume V et contenant N molécules de base) étant égal au nombre total
de degrés de liberté orientationnelle associés aux molécules (soit
2N), on aura la condition:

(Jq = Z N (18)
Bu 3
ou encore

dq - 16irX (19)
a=l *

où nv est le nombre de molécules par unité de volume avec n v = Nap/M

pour un échantillon pur, p étant la densité, M la masse moléculaire de
l'objet uniaxial utilisé comme molécule de base et N3 le nombre
d'Avogadro.
Dans le but de limiter le nombre de paramètres, on simplifie le
problème en utilisant l'approximation sphérique à une constante qui
considère que i) une seule viscosité -n et une seule constante
élastique K interviennent, ii) on ne fait aucune distinction entre les
modes splay-bend et twist-bend et iii) on ne fait pas de distinction
entre les modes se propageant parallèlement et perpendiculairement au
directeur. On a donc un unique vecteur d'onde de coupure qc et on
obtient alors

erf(—)V2
_i/2 kRT q_ Vnc '
<nx(0> nx(T}>- — !--,I . (20)
•n2 K
(JL)i/Z
T
c

ou on a

(21)

Dans le cadre de l'approximation sphérique à une constante, le

paramètre d'ordre S associé au désordre (uniaxial) dû aux modes
élastiques et (19) s'expriment alors

S = I-
>
P et
1 q3 = 6 ^ (22b)
»

49
 36

A ce niveau de simplification ne restent que deux paramètres

indépendants. C'est cette approximation qui sera utilisée par la suite
dans les analyses de spectres expérimentaux et on retiendra le
paramètre d'ordre S et le temps de relaxation de coupure TC comme
tels.

Reste le problème du choix de l'objet uniaxial considéré dans ces

calculs.
Pour choisir cette molécule de base, il est nécessaire d'obtenir
des informations sur la structure, la conformation et le tenseur
d'ordre associés à la molécule dans des phases anisotropes
suffisamment fluides (c'est à dire où l'influence des modes lents est
réduite), près de la transition nématique-isotrope, par exemple, ou en
solution dans des cristaux liquides de faible masse moléculaire. Si le
tenseur d'ordre est uniaxe, alors la molécule idéalisée pourra être
identifié à la véritable molécule.
Si, par contre, le tenseur d'ordre s'avère ou est suspecté être
biaxe, la molécule réelle ne peut plus être identifiée à la molécule
idéale considérée dans la théorie. Cependant, et sous certaines
conditions, on peut encore utiliser le spectre de la véritable
molécule en tant que spectre de base. Ceci est possible si les
fluctuations à l'origine de la biaxialité de l'ordre sont suffisamment
rapides (vis à vis des temps RMN carctéristiques) et d'assez faible
amplitude.
Si l'on a résolu le problème du spectre de base associé à une
molécule, dans les cas uniaxe et, dans une moindre mesure, biaxe, la
question du choix de la molécule de base reste entière. En effet,
celle-ci à pour seule contrainte d'être un objet uniaxe, ce qui peut
être le cas de la molécule réelle même, mais des groupes de molécules
et le monodomaine nématique en entier répondent également à cette
définition. La taille de l'objet choisi détermine, selon (22b), le
vecteur d'onde de coupure qc et on a la loi d'échelle:

q~3 « T\fz « (1 - S)~3 tx n"1 (« M pour un système pur) (23)

Si le choix de la taille de la molécule de base et donc du

vecteur d'onde de coupure qc est a priori purement arbitraire, en
pratique on aura intérêt à choisir le plus grand qc possible,
c'est-à-dire la plus petite molécule de base (une molécule réelle dans
le cas uniaxe, un petit groupe dans le cas biaxe) afin de partir d'un
spectre de base le plus proche possible de celui de la molécule
réelle. D'autre part, ceci évite de diviser, de manière quelque peu
artificielle, les modes élastiques en modes rapides, internes à la
molécule de base, et lents.
y
t
 ..- **:

37

LIQUID CRYSTALS, 1993, VOL. 13, No. 5,645-665

Theory of NMR Une broadening by elastic modes in

nematic liquid crystals

by J. B. FERREIRAf, H. GERARD, D. GALLAND and F. VOLINO*t

DRFMC/SESAM/PCM-CENG-85X-38041 Grenoble Cedex, France

(Received 13 October 1992; accepted 12 January 1993)

The model proposed in [TJ to calculate the broadening effect produced by elastic
modes on NMR line shapes of nematic liquid crystals, in which all modes with
relaxation times longer than a NMR time scale AT are static, and infinitely fast in the
opposite case, is put on a more rigorous theoretical basis, by considering all modes
with their actual relaxation times. The correlation function of the transverse
magnetization is calculated and expressed in terms of the self-correlation function of
the components n, and n, of the local director, assumed to be equivalent and
independent gaussian random variables. Formal expressions are given for the
general case, and in the one constant, cylindrical and spherical cases, approxim-
ations. The general procedure describing how to use this formalism for a NMR
spectrum composed of many Unes, is given. This formalism is then used to analyse
the same data as in [7] concerning a main chain nematic polymer, in the spherical
approximation. It is shown that fits with the same quality are obtained. These
results provide (Q theoretical support for the model Of[T), (ii) an operational way to
define At and (Ui) a practical example for discussion of the controversial problem of
the 'cut-off wavevectoi(s)' of the modes, which define the size of the elementary
iiniaiial object in the nematic medium. It is shown that, for this polymer, the
smallest size corresponds to a volume between one and four repeat units. The
analysis of line shapes provides the viscoelastic parameter ijl'2/K3'2, where TJ and K
are the average viscosity and average elastic constant Values of n and K can be
deducedfromthe theory. The limitations of the model are discussed. It is shown that
the present model and the one of [7] are complementary. It is argued that the
present formalism may be useful to analyse NMR line shapes in conventional
polymers.

1. Introduction
The study of thermally induced long range orientational fluctuations (the elastic
modes) [1, 2] in uniaxial liquid crystals is currently the subject of considerable interest
The main experimental techniques used for this purpose are light scattering [1, 2] and
nuclear spin-lattice relaxation [3-6]. While the theory to analyse light scattering data is
well established and widely accepted, essentially because this very technique can select
a particular wavelength which is always much larger than a molecular length, the
situation is more confusing for spin-relaxation because all modes, in particular those
with short wavelengths, need to be taken into account In a recent paper [6 (a)], Faber

* Author for correspondence.

t On sabbatical leave from Centra de Fîsica da Matcria Condensada (CFMQ, 2 Av. Prof.
Gama Pinto, 1699 LJsboa Codex, Portugal.
t Member of CNRS, France.

0267-8292/93 SIOOO © 1993 Taylor & Fi Ltd

 a*-;

38

646 J. B. Ferreira et al

discusses a number of important questions on this matter, in particular the central open
question about what are the shortest (cut-on*) wavelengths (whose actual values are of
fundamental importance in the theory) below which the continuous theory breaks
down. His conclusion is that "the director field can be defined at any arbitrary length
scale" and consequently, that it is justified to assume that "all the misalignment of
molecules in a nen.atic can be described m icons oï director fluctuations, even on a
microscopic scale" [6 (a)].
Although this viewpoint is not widely accepted, the very existence of the elastic
modes is not questioned, and methods that are sensitive to them can be used to study
viscoelastic properties, test the theory and/or estimate the contribution of other
mechanisms to the measured quantity. Besides those already mentioned, another
possible method is the study of NMR (and ESR) line shapes. A reason why this method
is not widely used is that the overwhelming number of NMR studies have concerned
low molecular mass liquid crystals, in which the effect of these modes on the NMR line
shapes is negligible [4]. In other words, practically all modes are fast compared to the
relevant magnetic interactions. In polymer liquid crystals on the contrary, due to the
increase in the viscosities by several orders of magnitude, an appreciable fraction of the
elastic modes are slow, producing some 'static' broadening of the lines.
In a previous paper [7], the calculation of the corresponding effect was made by
introducing a NMR time scale AT and assuming that all modes whose relaxation time is
larger than AT are completely static, and infinitely fast in the opposite case. This
assumption naturally led to the introduction of the notion of'static order parameter1
Sn,,, which could be expressed in terms of an average viscosity ij, an average elastic
constant K, the NMR time AT and the temperature T. This parameter quantifies the
amount of orientational disorder introduced by the 'static' modes. This model accounts
very well for the experimental variation of the proton NMR line shape of a main chain
netnatic polymer monodomain with the angle between the axis of the monodomain and
the static magnetic field [T]. It also proved to be very useful to analyse the time
evolution of line shapes during magnetic reorientation of monodomains, from which
accurate values of viscoelastic coefficients could be deduced for a main chain [8] and a
side chain [9] nematic polymer.
This model applies only to cases where the static disorder is small (S,,,, S 0-9), its
main approximation lying in the partition of the modes into (completely) static and
(infinitely) fast ones. Moreover, the quantity Sn,, which is deduced from the
experiments is not an intrinsic property of the phase, since it depends on the NMR time
scale AT. Since in addition AT is a priori ill-defined for broad proton NMR spectra
(although this limitation is not too serious, since the dependence on this parameter is
only to the power 1/2X it would be very interesting to remove this approximation and
consider all the modes in the calculation of the line shape. The main purpose of this
paper is to present such calculations.
In §2, the general theory is presented and close mathematical forms for the
correlation function of the transverse magnetization are given, corresponding to the
cylindrical and spherical one constant approximations. In §3, we describe an
application of this theory to the analysis of proton NMR line shapes of a main chain
nematic polymer, and in §4 these results are compared to those obtained with the
approximate method of [T). In § 5, the question of length scale and cut-off wavevect ors
is discussed in the light of the results obtained with this particular polymer. In the
conclusion, the approximations of the model are discussed and the possibility of
application of the method to conventional polymers is outlined.

'}
 39

NMR line broadening by elastic modes 647

2. Theory
Zl. Generalformalismfor the case when the static magnetic field is parallel to the
mean director
We consider an idealized 'basic molecule' constituted by a segment picturing its
(cylindrical) symmetry axis, and oriented at an angle O with respect to the static
magnetic field. For simplicity, its NMR spectrum is assumed to be a single sharp line j
at a distance Wj=(Uj0P2(COs^) from the Larmor pulsation ct>0. This distance is a
maximum (=Wj0) when the molecule is aligned along the field. If the angle 0 fluctuates,
the Une shape will change according to the time scale and the magnitude of the
fluctuations. In the adiabatic approximation, i.e. when the fluctuations are fast
compared to the spin-lattice relaxation time, the line shape J/o>) of an ensemble of such
magnetically uncoupled molecules is given by the Fourier transform of the function
GjM [10]
G/t)=Re {exp(ia>0f-r/T2)<expDFj«]>}, (1)
where
P
I <ûll")dt'=<0io \
C'
(2)
Jo Jo
IfT2 is a damping coefficient which ensures that G(I) always tends to O at infinite time, or
equivalently, that the NMR line has a non-zero width (equal to 2/T2) in the absence of
fluctuations. The brackets stand for an equilibrium statistical average.
The calculation consists of evaluating the above quantities, assuming that the time
fluctuations of 0 are due to the elastic modes. We will adopt the viewpoint discussed in
the Introduction, and identify the local director n of the continuous theory with the
long ans of our idealized molecule.
In the hydrodynamic theory, a statistical order parameter S at time f is defined as

(3)

where B1 and «, are the components of the local director in the laboratory frame, z being
along the mean director H0 (Le. the axis of the monodomain) and where, here, the
brackets stand for a space average at time t'. Invoking the ergodicity theorem (valid in a
sufficiently fluid medium), it can also be thought of as a time average performed-at a
particular point r, where our idealized molecule is situated. Thus, Itx and n, are the
components, in the laboratory frame, of the unit vector along the molecular axis.
Consequently, PJ[COS 6(0] in equation (2) is just the value at time t' of the quantity
inside the brackets in equation (3). In the following, the explicit dependence on r will be
omitted for simplicity, in all formulae.
Defining the real random functions
'I (4)
*'! and

--fa* P (5)
we can write

(6)
 40

648 J. B. Femira et al.

and

2.2. Statistical behaviour X7(I)

Later, it will be assumed that njt") and njf) are gaussian random functions with the
same distribution, and it follows that n*+n| is a random function of the x1 type with
two degrees of freedom. But whatever reasonable hypothesis is made about the
probability distributions of Hx and H7, or {, X/t) defined by equation (S) may always be
considered as a gaussian random function. The reason is that an integral of a random
function behaves just like the sum of random variables, and the gaussian character
follows from the central limit theorem. This argument is probably weak when t is small
or comparable to the correlation time of {, but in this case Xff) is always small, and the
type of distribution taken for this variable is not likely to affect much the average
present in equation (T).
With this gaussian hypothesis for Xf we have

<exp[Ur/t)]>= (8)

where

P(XpI)* (9)
Noting that <X^> =0 (this property follows immediately from equations (4) and (S)) we
readily obtain
<exp[iX/t)]>=«p[-aJ(t)>/2]. (10)
Assuming that £(f) is a stationary random function, which is actually the case whenever
S does not depend ontime,it can be shown after some mathematical manipulation that
we have

where CJ(T) is the self-correlation function of (

(12)

2.3. An expression for CJ(T)

From equation (4), it follows that
(13)
where we have used the property that <i£+nj> does not depend on f and t. Using the
fact that H. and n, are equivalent random variables (same statistical properties), we
obtain
(14)

V
\
 41

WMR line broadening by elastic modes 649

To go further with the calculation, we shall follow [S] and assume that (i) the
fluctuations in x and y are uncorrelated and (ii) nx and n, are gaussian random variables
with the same distribution. The fust assumption makes it possible to write

and the second assumption leads to

Inserting these two relations into equation (14), we arrive at the following simple result:
t)>2. (15)

2.4. Formalism for the case when the static magnetic field is at an angle Q to the
mean director
The laboratory frame is chosen such that the z axis is along H0, as before, and the
static field H is in the xz plane. In this frame, the polar and azùnuthal angles of the field
and of the molecular axis n are 6, 0, and Q, 9, respectively. The relevant formulae are
still as in equations (1) and (2), but now 0=(n,ff) becomes a function of 6. We have

Pï(cosfl)=f(i^sinïe+Hïcosî0+2iixnIsinôcose)-i (16)
and taking <i£>=<i£+n?>/2, <iyi,>=0, it follows that
<P2(cos0)>=SPa(cose). (17)
Writing o>XO=<<i>/0>+[«»j(0-<(B/tO>], and using equation (17), we arrive at
tojM^SUjoPAcasey-frjoteO, (18)
where
(19)
is a random function which generalizes ftf), given by equation (4), to the case where
e#0, Le. £0(f)3«t), and also satisfies <&(O>=0.
Similarly, denning the random variable

*./<)=-K, r Jo
(20)

Vl the lineshape /«./to) in the presence of fluctuations, when the director is rotated at an
angle 6 with respect to the field, is given by the Fourier transform of Ge,/t)

(21)

Invoking the same arguments as for X ft), we shall assume that X». /O is a gaussian
variable, and we can write a set of equations similar to equation (8) through equation
(12), the only difference being that X ft), CJ(T) and {(f) are now replaced by Xt ft),
 42

650 J. B. Ferreira et al.

The calculation of the self-correlation function C4Jt) is made in three steps. First,
using equation (19), and all statistical properties of nx and n, previously assumed to
hold, we obtain

-4sin 0 cos ©

+4sin ©cos3 ©<n2(iX(t'-tK(t'-T)>. (22)

This equation is exact, but unuseful because most of its bracket expressions are not
expressable in terms of model parameters such as S or the viscoelastic parameters
referred to below. To proceed, we assume that the fluctuations are small (nj and n2 « 1);
then we can replace nz by 1-itô+n2). Noting that <nJ(t')) = <n2(t')n*(z'-T)>
t'—T))=0, and neglecting terms of the third order in n2, we can write

(23)
In the Appendix, we show that the gaussian hypothesis for njf) leads to the relation
<^(r>^f-t)>=3<n2><nJt(C)nJ1(f-t)> (24)
so we find

T)>2- (25)
For 0=0, we recover all equations of the previous sections.
It is interesting to note that, in the general case, the correlation function of £e is the
*î* sum of two terms, the first one being proportional to the correlation function of n, and
ï the second one to its square. For 0=0 and 90° on the contrary, only the square term
survives. Because this latter function is basically a quantity (much) smaller than 1, and
the coefficients of proportionality are usually comparable when 0 is around 45°
(otherwise the coefficient of the square term largely dominates), we conclude that the
effects of the elastic modes GB NMR line shapes are stronger when the director of the
nemat:= monodcmain makes a large angie (say, 40 to £0°) with the static field. This
result constitutes the theoretical support for the qualitative arguments invoked in [7]
to justify the corresponding expérimenta! results. The only thing that is now needed to
perform a practical calculation is a formal expression for the self-correlation function of
nx in terms of the parameters of the hydrodynamic theory.
1
 43

NMR line broadening by elastic modes 651

15. Calculation of <njit')njif-t)y
In the hydrodynamic theory of nematics [1,2], we inboduce the space Fourier
transform of na limited to the volume Kof the ntmatic monodomain

I dqnjiq, t)exp(-ig.r). (26)

Using the normal splay-bend («=!) and twist-bend (ot=2) modes njiq,t), the
correlation function of R1 can be expressed as

showing that the two kinds of modes contribute additively to this function. Neglecting
the diamagnetic contribution to the free energy (see below), the <K(g,0)|2> andthe t^q)
can be written [2]

(28)

where
qjp4—

«! (296)

In these expressions, the symbols have their usual meanings. The K, are the Frank
elastic constants, the B1 are the Leslie viscosity coefficients and the ^ are the Miesowicz
viscosity coefficients. The quantities ^x and qz are the components of q perpendicular
and parallel to the director z, and p*=qjqr
Introducing the two functions
(30)
and
(3D
equation (27) can be rewritten

(32)

This equation gives the most general expression for the correlation function OfRx. The
value of this function for T=O is
(33a)
where £, is a quantity, with the dimension of an energy, given by

—=— f (336)
Ee 8jt3«=i
This energy may be thought of as an elementary elastic energy associated with the
f
modes, and stored in the nematic phase. It must be sufficiently large compared to the

\
 44

652 J. B. Fcrreira et al
•quantum' of thermal energy fcBT/2 so that a nematic phase exists at temperature 7! The
energy Ec is related to the order parameter S via the evident relation
3fcBT
S=I- (34)

So far, nothing has been assumed concerning the molecules, except that they are
necessarily objects with cylindrical symmetry. If we assume that the nematic domain
contains JV such objects, a condition concerning the integration volume in q space can
be written. This condition is that the number of modes in this monodomain is equal to
the total number of orientational degrees of freedom associated with the (linear)
molecules, namely 2N [6J Noting that the density in q space, of modes of polarization a
(a =1,2), in a sample of volume V, is K/8n3, the above condition writes

i*»ii jU- 2N. (35)

Introducing the number of molecules per unit volume n,, the density p, the 'molecular
mass' M (the quotes emphasize the fact that the idealized molecules are not necessarily
the true molecules) and the Avogadro number N1, this condition can be rewritten

£ JiIg=WJi3B, (=l6rt3jv«^ for a pure sample J. (36)

The integration volume in q space has necessarily cylindrical symmetry around the
mean director. The lower limit corresponds to the inverse of the sample dimensions. It
can be taken as zero for a macroscopic sample, in normal magnetic fields, provided the
upper limit is sufficiently high (of the order of the inverse of the molecular dimensions),
as is actually the case. The reason is that the contribution to the integrals in equation
(27), of all modes whose wavelengths are longer than the magnetic coherence length (a
few microns in our magneticfield),turns out to be completely negligible. This result also
justifies that the diamagnetic term in the bee energy can be neglected, as previously
assumed. The upper limit of the integration volume is more interesting. It cannot
extend to infinity in all directions since equation (36) imposes that this volume be finite.
The upper limit of the integration volume thus defines two cut-off wavevectors gj, for
each orientation O of q with respect to the mean director.
To proceed with the calculation, some volume shape should be assumed. We choose
the simplest shapes, namely circular cylinders. In this case, there are, in all, four cut-on*
parameters, q^ and (f^. The volume element in q space being dq=2nq±dqidqr,
equation (32) can be rewritten:

The integral in q± being performed at constant qv we have dq± = q, dp. Noting that the
integrand is an even function of ^1, the change of variables (q»qj-»(p=qjq,,v=qj
leads to

(38)
 45

JVAfR line broadening by elastic modes 653

where the JJt) are double integrals defined as

'pa> f J»exp[-t)*tF«(p)]. (39)
10/^P)J o
The upper limit of the second integral is given by:
9.Jf)=<&* (4Oa)
and
(405)
Introducing the error function erf, which verifies the identity

ir
>ij0 (41)

we get

.
(42)

It can be easily seen that both integrals converge for all t values, but their evaluation
can be made only numerically. However, for T=O, a close form exists. One way to
perform the calculation is to put T=O in equation (32) and follow the procedure used in
[TJ. The final result, expressed in terms of the elementary elastic energy £e of equation
(33Xis

(43fl)
where

a=l,2. (43 fc)

It is worth noting that, from the purely computational point of view, the model
implies eleven parameters, namely the five independent viscosity coefficients, the three
elastic constants, and the four cut-off wavevectors related by equation (36). This is
certainly too much in most practical cases, where little is known about the viscoelastic
constants. Moreover, the numerical evaluation of the J1 for finite values of T is rather
difficult Some further simplification should thus be made.

2.5.1. The cylindrical, one constant approximation

The next simplification that can be made corresponds to the one constant
approximation, in which only one viscosity (17) and one elastic constant (JQ are
Vi considered, and no distinction between the splay-bend and twist-bend modes is made,
so that the index OE becomes irrelevant. Defining perpendicular and parallel 'cut-off
relaxation times' T^ and t£ as

(44)

:)
 3F-

46

654 J. B. Ferreira et al.

we obtain, after a rather lengthy algebraical manipulation, the following final (exact in
this approximation) result:
Tt
t. T
—.1/2 ii ftgl e
PŒ>
I
(45>
2 4Tt2 K *JT/ll
This integral is easily evaluated numerically for any set of values of the two viscoelastic
parameters and the two cut-off wavevectors. In this approximation, equations (43) and
(36) become

and
<«•>
4»i*fl.T? for a pure sample ). (46b)
M J
In this cylindrical approximation, the number of parameters is reduced from eleven to
three. It is possible to simplify the problem slightly further with the spherical
approximation.

2.5.2. The spherical, one constant approximation

This is the usual approximation that is found in the literature [4-6]. No distinction
is now made between modes propagating parallel and perpendicular to the director.
The integral in equation (36) is calculated in a sphere whose radius is the modulus qe of
the single cut-off wavevector. The result of this calculation is well known [5]
1 2 2
IT
ft ' I KB'
1 t T erfÏT/TCJ
VH\T/T V tMrn
(47)

The single cut-off relaxation time TC is given by

(48)

and equations (43 a) and (36) become

(49a)

and

6s2N«^ for a pure sample). (496)

M j
4
^ The two free parameters of the model, to be determined by comparison with

"1 experiment, are the cut-off relaxation time TC and the order parameter S. In terms of
these parameters, equations (47) and (34) can be rewritten

(50)
and
S=I-
3fcBr I?172 (51)
 47

line broadening by elastic modes 655

These two equations are the central ones for practical use in the framework of the
spherical approximation. They will be used below to analyse proton Ntef line shapes
of a nematic polymer. But, before, we shall summarize the procedure to be followed to
perform the complete calculation for an aligned NMR spectrum (basic spectrum)
constituted by a single line, and then for a more realistic basic spectrum composed of
many lines.
2.6. The calculation procedure
In short, the calculation of the NMR line shape in the presence of elastic modes,
when the director is rotated at an angle 0 to the static field, for a basic spectrum
composed of a single line at Oj0, can be made in six or seven steps:
(i) choose an expression for the correlation function of njtf), namely either
equation (32) (general case), or (45) (cylindrical, one constant approximation),
of (47) (spherical, one constant approximation);
(ii) select numerical values for the parameters of the model and the angle O;
Ou) calculate the correlation function of nx for all t values between O and about
5T2;
(iv) insert these values into equation (25) to calculate the correlation function of
t»
(v) use equation (11) to calculate <X| /t)>, and equation (10) to calculate
<exp[iXe./t)]>;
(vi) use equation (7) to calculate G01XO. and perform the time Fourier transform,
(vii) Possibly, if the sample is a polydomain, with a distribution function F(O, $)
for the mean directors, the relaxation function [G/r)],, to be Fourier
transformed is

(52)

The isotropic "powder' corresponds to F(0, "

2.7. Generalization to complex spectra of real molecules

So far, our molecules are idealized objects with cylindrical symmetry which are
reduced to their long axis. Real systems are composed of real molecules which generally
have no symmetry at all, and their NMR spectra are composed of many lines. A central
question to be discussed is which object should be identified with the idealized molecule
of the theory. Whatever this object is, we shall suppose that the perfectly aligned (basic)
spectrum of the idealized molecule is the same as that of the real molecule (this
assumption will be discussed below).
To know this spectrum, it is generally necessary to conduct experiments using
anisotropic fluid phases in order to determine the structure, conformation and order
tensor. The principal axes OXTfZ of this tensor are the most reasonable choice for the
molecular axes. For molecules as large as those of usual low molecular mass nematics, it
Vi is expected that these axes do not change much with temperature and with the medium
in which they are embedded, so that they can reasonably be considered as rigidly
attached to the molecule. If such experimental results are not available, then the
directions of these axes, the structure and conformation should be guessed.
Two possibilities arise, depending on whether the order tensor is uniaxial or biaxial.
If it is uniaxial, the basic spectrum of the idealized molecule can be identified with that

\
 48

656 J. B. Fcrreira et al...

of the real molecule, calculated with OZ r-'ong the field, using magnetic interactions
that are averaged over internal motions only. These averages are perfectly denned in
the framework of the single conformation model [12,13].
If the tensor is (suspected to be) biaxial, we may proceed as follows. The overall
motion described by the order tensor is split into a fluctuation (described by angle jf) of
principal axes OX and OZ in their own plane (Jf and Ydefined such as Sn-Sn SsO),
and a uniaxial motion of the frame defined by OX', OY and OZ', where OJf', OY and
OZ' arc the average orientations of OX, 07 and OZ over the fluctuations *. The
magnitude of this fluctuation is related to the anisotropy of the orientational order

~ « . r» . ft e* 1 ^O I" V^/
i+ns
The uniaxial order parameter to be associated with the long axis OZ' is related to the
principal values of the order tensor by

For uniaxial order, we recover Z=O and S=S21.

The basic spectrum l(ca) to be considered is now the one calculated with OZ' along
the field, using the magnetic interactions that are partially averaged by the internal
motions, as in the uniaxial case, but also by the fluctuations %. This procedure, which
applies under the assumption that the fluctuations x are fast but not too large
(IrU <*/4X and uncorrdated with the fluctuations described by the elastic modes, has
been briefly justified in two previous papers [12,13]. Th: paper presenting the detailed
calculations is in preparation.
It is also worth mentioning that the real molecules may (but not necessarily) be
identified with the idealized molecules if their orientational order can be considered as
unîarâl However, strictly speaking, they can definitely not, if the order is biaxial. Since
thite is increasing evidence that the molecular order is generally biaxial in real systems,
this means that the idealized molecules of the theory should necessarily be identified
with entities larger than one molecule.
Let J(o>) be the basic NMR spectrum of one molecule, calculated with the field along
OZ', according to the above prescription. This spectrum is composed of L lines j of
intensities I1 and distances (Oj0 from the Larmor pulsation O0. We can write

/(o>)= £ J^OIj0-W0). (55)

J=i
TherotationaroundOZ'isunifonnbydefinition.Ifitcanbeconsideredfastenoughon
AeNMR time scak, all the averaged (by the internal motions and the uniform rotation)
^ magnetic interactions arc scaled by the factor P2(cosfl). The rotated spectra have
•: exactly the same shape as the aligned spectrum, but are narrower by this factor. In this
! case (and, strictly speaking, only in this case), each line j keeps its identity when the
angle is changed, and can thus be considered as independent of all the other lines.
. ' Now, if the molecule is embedded in a nematic phase, it is evident that, because of
• this independence, the lineshape /4(01) in the presence of elastic modes is the time
Fourier transform of G^t) given by

= .l¥W>> (56)

64
 49

NMR line broadening by elastic modes 657

where Ce,/t) is the relaxation function associated with linej of the theoretical spectrum,
given by equation (21).
In all this discussion, it is assumed that the fluctuations x and the rotation around
OZ' are fast compared to the largest magnetic interactions. If they are not, equation (56)
does not apply, although it may be a reasonable approximation if the amplitude of* is
small and the order sufficiently large.

3. Application to a nenutk polymer

The theory is now applied to the analysis of the proton NMR line shape of a main
chain nematic polymer of the type (RF)1, where R is a mesogenic unit and F a flexible
spacer, labelled AZA9dl4, whose chemical formula is

CH, CH,
This sample is the same as the one used in previous work [7, 83, in particular, the work
in which the concept of 'static order parameter' was introduced [7]. Here, we repeat the
analysis of the same data, namely the angular (Q) dependence of the proton spectrum of
a monodomain of this polymer, at 393 K, using the present theory. For this analysis, we
use the spherical, one constant approximation, which will prove to be sufficient for our
purpose.
According to §2.7, we need as a starting point the theoretical proton NMR
spectrum of the idealized basic molecule, with the long axis along the field (basic
spectrum). For polymer AZA9dl4 (deuteriated on the spacer), we assume that this
spectrum is the same as that of its azoxybenzene moiety. The exact simulation of the
corresponding twelve spin 1/2 spectrum has been the subject of previous work, recently
published [14], in which the structure and conformation of the azoxybenzene moiety,
and the direction of the principal axis OZ and associated principal value Szz of the i
order tensor, were determined. Although there is some independent evidence that the
order is biaxial [ISJ, no significant information about the biaxiaiity can be extracted
from this simulation alone, because the effect of introducing a finite value of ijs can be
very easily compensated by slightly changing the geometrical parameters [14]. ,
However, because the existence of biaxiaiity also changes the width of the aligned '
(S= 1) spectrum, two basic spectra will be considered, one for Ij5=O, and one (almost
identical, but narrower, since some dipolar interactions have been reduced by the z
fluctuations) for ^=0-272 (corresponding to IjL= 22-17°), which is probably closer to
the actual situation, as shown by a recent, more detailed, analysis (performed in our ^
laboratory) of the whole set of existing data concerning this polymer. These two basic )<
spectra are reproduced infigures1(A) and (B). |
The order parameter S22 of AZA9dl4 is known over the whole nematic range ,
[7, 14} At 393K, its value is »0-54. The values of the uniaxial parameter S of the M
theory,givenbyequation(54),areS=0-54withbasicspectruml(A)andS=0-69with .'
basic spectrum 1(B). In the spherical, one constant approximation, the single
remaining parameter to be determined from the analysis is the cut-off relaxation
time T^
We have performed the calculation as prescribed in §2.6 for the same values of the >'
angles as in [7], using a UNIX computer. The enormous number of lines of the .r
1
«heoreticalspectrahavebeengatheredintoSOOpacketsofwidthAto^aroundujp.The I
 50

6SS J. B. Ferreira et al.

D
. 10 kHz.
Figure 1. Basic proton NMR spectra of AZA9dl4 polymer used in the analysis. (A) calculated
for no biaxiality Ob=O): this spectrum is identical to spectrum 1 (fc) of [14]; (B) calculated
with »is=0-272: this spectrum is narrower than (A) by the factor (1 +ifc)~ *; (Q calculated
for a large basic molecule such that 5=0-962: this spectrum is narrower than (A) by a factor
0-54/0-962; (D) reference spectrum used in the approximate method of (TJ: this is the
experimental aligned (6=0°) spectrum at T= 12O0C for the nematic phase.

intensity I1 associated with pockety is the sum of the corresponding intensities of the
individual lines inside this packet, and this packet is identified with line j of the
spectrum used in the computation.
The results of the simulation, with T2=O1ISxIO-2S, are shown in figures
2(A) and (B) for the two basic spectra. Comparison with the experimental spectra (see
figure 2(E)) shows that good fits, of nearly equal qualities, are obtained in both cases.
For the case I)5=O, the best fit is obtained for TC=(0-43 ±0-1S) x 10~6 s. The simulation
is rather sensitive in the sense that values of tc outside the range mentioned can be
excluded. The situation is slightly different for the case i;s=0-272, where the uncertainty
ft range is broader, although the central value is certainly larger than for the case fs=0.
Figure 2(B) corresponds to the simulation with rc=O95 x 10~6 s.
Introducing in equation (S 1) the values of S and rc for the case ijs=O yields ijl/2/K3'2
Vi =(M3±0-2!)x lO'kP'^dyne-3'2 (IkP=IO 2 Pas; ldyne=MTsN). Exactly the
.. j same value is obtained for the case r;s=0-272. This quantity is the only information on
the viscoelastic parameters that can be extracted from the simulation of the line shapes.
It is possible to go further and find the actual values of n and K, using equation
(49 b). Consider the case ijs=0. The smallest uniaxial object is then the repeat unit (the
object attached to the OXYZ frame of §17). The corresponding molecular mass is
426g. With p=lgon~ 3 , equation (49*) yields 9c=4-37xl07cm"1. Combining
equations (34) and (49 a), we obtain K=0-78 x 10~6 dyne, and from the above value of
tllaIK31*, we deduce ij=(a61±022)kP.
 51

NMK line broadening by elastic modes 659

_/wi j*\A.
D

f
Figure Z 90 MHz proton NMR spectra of a nematic monodomain of AZA9414 polymer, at
393 K, for different angles 6 between the mean director and the static magneticfield:(a) 0°,
(t>) 25-5°; (c) 37°; (<f) 47°; (e) 76 5°; (/) 90°. Spectra (A), (B) and (Q are calculated in the
spherical, one constant approximation: (A) using basic spectrum 1(A), 5=0-54 and
T1=(MSxUT** (case uniaxial); (B) using basic spectrum 1(BX 5=0-69 and T0=035
xlO~ 6 s (case biaxial); (Q using basic spectrum 1(Q, 5=0-962 and tc=0-64x!0~*s
(large bask molecule). Spectra (D) are calculated with the approximate method of [7],
using reference spectrum 1(D) and 5^=0962 (reproduced from [T)). Spectra (E) are
experimental, obtained at T= 12O0C for the nematic phase (reproduced from [T]).
 52

660 J. B. Ferreira et al.

It is instructive to compare these values with the results from magnetic reorient-
ation measurements [8]. Concerning the elastic constants, il has been found [8] that
JC, «s 1 x HT'dyne and that 0-3sS/C3/K, «0-5. Nothing is known about K2. If we
assume that the value of the latter is comparable to that oi the two other constants, it is
seen that the value obtained for K may be considered as a reasonable average value.
The agreement is better (more accurate) for the viscosities, which can be classified into
small and large [83, differing by about two orders of magnitude. The small ones are
TbO1* Ib and probably IJB and the large ones areiflwU, (=y,), ij.pi.y and tjc. The viscosities
with indices bend, twist and splay are given by equations (29). They correspond to those
of the three fundamental modes, and are equal to Ij1(O) (=I2(O)). fj(<») and 1i(°°),
respectively. For the present purpose, it is natural to focus on the viscosities of the three
fundamental modes, whose values are (in kP) FJB]: i;bend=O-17 ± 0-02, ^11, = 11 ± 5, and
Ij0Jj1,=15-7. The harmonic average (average of the inverses) of these three viscosities is
ifh.rm=(0-50±OO7)kP. It is remarkable that n^m and ij, which have been obtained
independently from one another, are found to be equal within experimental accuracy.
In a situation in which the three elastic constants are of the same order of magnitude,
the harmonic average of the three fundamental viscosities is proportional to the
(arithmetic) average of the relaxation rates of the three fundamental modes. And it is
clearly this average rate which is the most reasonable candidate to be identified with the
single average rate of a one constant approximation model.
The fact that the simplest possible version of the model (one constant, spherical),
with one free parameter only (rj, allows us to describe satisfactorily the experimental
results, and that the values deduced for the viscoelastic parameters are in complete
agreement with what is expected, suggests that the physics introduced in the theory (the
elastic modes) is probably correct. However, the analysis is heavy and the amount of
information about the viscoelastic properties extracted from the analysis of the line
shapes (the value of i71/2/K3'2), is rather weak. From this point of view, it may be asked if
the simpler method Of[T], when applicable, does not give in fact the same amount of
information. This is what is explored in the next section.

4. Comparison with the approximate method of [7]

As stated in the Introduction, the approximate-method described in [7] introduces
a NMR time scale AT which is not only a function of the system under study, but also of
its NMR features. In this description, the disorder associated with the slow modes is
characterized by a 'static order parameter' Sn., given by the same expression (Sl), but
with te replaced by At [T].
An important question in this method is how to define accurately AT when dealing
with broad (proton) NMR spectra, such as that of our polymer. All that could a priori
be said is that "AT is of the order of the inverse of the spectral width" [T]. This statement
can now be made much more accurate using the above results. Combining the two
expressions for 5 and £„.,, we obtain

(57)
With the above values of S and te associated with either i;s=0 or 0-272, and
$«.,=0-962 [TJ, we obtain AT=(0-63+0-22) x KT*s, which is about the value 10"*s
guessed in [TJ
It is interesting to relate AT to the full extension Avref of the reference spectrum of
figure 1(D). This spectrum extends over ~ 15 kHz, corresponding to (Aw,,,)"1
 53

NMK line broadening by elastic modes 661

=(2*Av tcf )~ I ~H x 10~ss~0-17At. This result means that, for broad NMR lines
such as those of the AZA9414 polymer, the value of At to be chosen, in an analysis in -
terms of static order parameter, is about 6 (Ao>Icf)~ '. Thus, with this prescription for the
choice of At, the approximate method of [7], when applicable, yields the same
viscoelastic information as the more rigorous method presented here. This fact justifies
a posteriori its use in the analysis of line shapes in magnetic reorientation experiments
[8,9]-

5. The problems of Oe cat-off wavevector and of the nature of the basic molecule
In this section, we use the above results to discuss the questions of the value of the
cut-off wavevector and the nature of the basic molecule in the AZA9dl4 polymer, in
order to see if it is possible to shed new light on this controversial [6 a] matter.
In a matter like this, the most important point is to separate properly the quantities
introduced in the model into quantities with direct physical meaning, that is, which are
directly accessible to experiment, and those which are only convenient concepts
introduced in the theory. The order tensor of the mesogenic unit (which can be
measured directly from simulation of the NMR spectrum, if the structure and
conformation are known), the viscosities and elastic constants (which both can, in
principle, be measured directly by well-established methods, independent of NMRX
and the density belong to the first class. The cut-off wavevector(sX the cut-off
correlation tiros(s) I0 the uniaxial order parameter S, the basic molecule and its
associated theoretical NMR spectrum, on the contrary, belong to the second class.
These latter parameters can a priori have any values, the only restrictions being that
calculation of the measurable quantities in terms of them should always yield the same
results. The restrictions on S, t0 qc and n, (or M) can be written in the form

" T""2 , (58)

(59)

qlnJ * = 6n2 (or qlM=6n*pNf for a pure system). (60)

These equations show that the choice of the cut-off (and thus of the associated basic
molecule) is almost completely arbitrary. The smallest possible basic molecule
corresponds to the smallest entity in the nematic monodomain which can be
considered as having uniaxial symmetry, and the upper limit is the size of the sample
monodomain.
This statement can be made more clear if we remember the definition of the
theoretical NMR spectrum of the basic molecule: it is the spectrum of this molecule,
VT- assuming that its symmetry axis is fixed along the field. However, in order to calculate
'! this spectrum, we have to consider the real magnetic interactions iu the real molecular
units inside this basic molecule, which are averaged by the real molecular motions. The
larger the basic molecule, the larger the number of motions which need to be considered
(in particular if the size is large, some short wavelength modes should be considered as
internal motions). The more and irore averaged the magnetic interactions are, the
narrower will be the basic NMR specrrum. The difference to the experimental spectrum
decreases, and this is equivalent to an increase in the uniaxial order parameter S. For a
basic molecule equal to the monodomain, all motions are internal motions, the basic

\
 54

662 J. B. Ferrsira et al.

spectrum is i^sntical to the experimental one, and S= 1. Correlatively, the slower the
motions associated with the (remaining) modes to produce the observed broadening,
the larger tc. For S-* 1, T0-* oo in agreement with equation (58) and, according to
equations (59) and (60), qc-»0 and nn-»0 (or M-* oo). To summarize, wj have the
following scaling laws (valid in the spherical, one constant approximation).
1
(or oc Af for a pure system). (61)
This statement is consistent with the results obtained for our "vjlymer. We have
indeed seen above that, with the introduction of the biaxiality, which is equivalent to
including more disorder inside the basic molecule (and thus less disorder in the modes),
an almost equivalent fit is obtained, as with no biaxiality, but with a larger value of TC.
According to the above scaling laws, the basic molecule, which, in the uniaxial case,
corresponds to one repeat unit, corresponds now to ~3-3 repeat units. The basic
molecule is now larger because each unit performs its biaxial fluctuations inside the
uniaxial frame OJKTZ' (cf. §2.7). This frame is necessarily attache ! to an entity larger
than one repeat unit Let us now increase the scale by a large factor, and assume that the
internal motions include a very large number of modes. The basic NMR spectrum is
now much narrower, almost equal but not quite, to the experimental spectrum. This
means that the order parameter S associated with the remaining modes is very large.
Suppose we choose for S the value of Sn,, =0-962, According to the scaling laws, the
associated value of TC is now increased to 063 x 10~* s, qf is reduced to 5-29 x 106 cm ~ *
and M increased to ~7-6 x 105 g. The basic molecule of the theory corresponds to a
volume of ~(1(X)A)3, containing ~1800 repeat units. Figure 2(C) shows tàe
corresponding calculated spectra. It is seen that, although still reasonable in the sense
that the main features are reproduced, the fit obtained is significantly worse than before
(see figures 2 (A) and (B)X and also worse than the fit obtained with the analysis in terms
of 'static order parameter' (see figure 2 (D)): the spectra are too well-resolved compared
to the experimental spectra. The reason for this is clear, the basic NMR spectrum (see
figure 1 (Q) has been deduced from spectrum 1 (A), assuming that all the internal
motions are infinitely fast With a basic molecule as large as that chosen here, the modes
included as internal motions are not infinitely fast, and consequently the basic NMR
spectrum is too well-resolved. This illustrates the practical limitation of the choice of a
small qc in the present method, and correctively the power of the appropriate method
of [73 for this particular system.
It is seen that, in the problem of the broadening of NMR line shapes by the elastic
modes, although the cut-off wavevector is in principle arbitrary, in practice it is not. In
practice, we must choose as the basic molecule the smallest possible object which can
reasonably be considered to have the uniaxial symmetry, in order to avoid the problem
of slow internal motions. This object is one single molecule (in the case of uniaxiality)
or, more likely, a small group of molecules (in the case of biaxiality). On the other hand,
even if there is no problem with slow internal motions, it is not reasonable to choose too
large a basic molecule. The reason is that it is not natural to split the modes arbitrarily
into short wavelength modes, which would be considered as internal motions, and long
wavelength modes, which wc~id be the 'true' elastic modes. Unless there is some good
reason for not doing so, all modes must be considered as 'true* modes. This is another
argument favouring thechoicejor the basic molecule, of a small (molecular size) object
with uniaxial symmetry. It follows from this discussion that, in practice, the choice is
rather limited. Only careful analysis of relevant experimental results can tell what is the
1
'best* basic molecule in each particular case. We have seen that for our polymer, the best
I

P
 55

Une broadening by elastic modes 663

results ai • . ^ aned with an object whose size ranges between one to four repeat units.
Similar situations are suggested by spin-lattice relaxation results in several systems [4-
6]. All these results are consistent with Faber's view that "all the misalignment of
molecules in a nematic can be described in terms of director fluctuations, even on a
microscopic scale" [6 (a)], keeping in mind that the microscopic scale is probably, in
real systems, larger than one molecule; larger in order to account for the probable
biaxial motions which, by nature, cannot be described by these fluctuations.

6. Concluding remarks
To conclude, we wish to make some remarks on two points, namely the limits to the
applicability of this theory, and the possibility of its use to study viscoelastic properties
of soft, isotropic materials such as conventional polymers.
The limits of the theory are associated with the several unavoidable simplifications
that have been made, mainly (i) the assumption of small amplitude fluctuations, (ii) the
gaussian statistics and the non-correlation between n, and nr and (iii) the gaussian
character of Xft). First, the small amplitudes are essential approximations of the
hydrodynamic theory which leads to equations (26) through (29). It turns out that the
most severe numerical approximation in all the calculation lies in the identification of
the overall mean square amplitude fluctuation <0Z> to (sin2 0>: the relative error made
is of the order of 0*/3. On the other hand, equation (25) suggests that S cannot be
smaller than 1/7=0-14 Other considerations, not developed here, show that this
limit is in fact 1/4=0-25. Suppose we fix the lower limit of S to ~ 0-429 (the lowest value
of the Maier-Saupe theory). This value corresponds to an average angle of ~ 38°, that is
~-0-66rad, yielding 0Z/3~15 per cent Second, the gaussian character is used to
establish equations (15) and (23). In fact, since the values of nz and n, are limited to the
range [ — 1, +1], the statistics cannot be exactly gaussian, so that the two equations are
exact only in the limit S=I. To have an estimation of the error made, we have
calculated the ratio R=<i£>/3<nj> as a function of S, using a gaussian distribution
function truncated at +1. We find that R decreases from 1 to 0-84 when S decreases
from 1 to 0-429. If the truncation is made at ±0-85, to take into account the (small)
correlation introduced by the condition i£4-nj < 1, the lower limit decreases from 0-84
to 0-78. It is seen that the error made here is of the order ~20 per cent
An overall precision of ~ 15-20 per cent in this problem is quite reasonable in view
of all the simplifying assumptions that were made, and the accuracy in the
measurement of the viscoelastic parameters, in particular in polymeric systems. Thus,
we can reasonably expect that the present description may be useful for systems in
which the degree of (dynamical) disorder corresponds to S as small as 0-4-0-5, that is for
all liquid crystals. It is worth noting however that this problem of a small value of S is in
fact not too serious, since S can be increased by choosing a larger basic molecule, whose
size, we have seen, is somewhat arbitrary.
Third, the approximation concerning the gaussian character of Xjit), pointed out in
§2-2, is a good one only for t sufficiently larger than the correlation time of £. This
means that this correlation time should be short This condition limits, in practice, the
applicability of the method to samples with average viscosities that are not too high.
There is however a way to escape this limitation, namely the approximate method in
terms of'static order parameter' [7], since no such condition is invoked. In the range
where it applies (Sm sufficiently large), this method is (or rather has become since, now
we know how to choose accurately the NMR time scale AT) as accurate as the more
 56

664 J. B. Fetteira étal.

detailed method presented in this paper, to extract viscoelastic information from the
NMR line shapes. :
This discussion shows that the two methods are in fact complementary. This
method presented here is rather suitable to analyse systems that are not too viscous,
with possibily rather small order parameters, while the method of [7] is adapted to
more viscous and more ordered systems.
As a possible extension of this work, we would like to point out that, although
developed for (polymer) liquid crystals, the formalism presented may also be useful to
analyse NMR Une shapes of conventional polymers. The reason for this is the
following. In conventional, rather soft (amorphous, melts, solutions, gels . . . ) polymeric
systems, there is certainly no long range order at a macroscopic scale as in nematic
systems, but some order at a mesoscopic scale (larger than, say ~ 100 A) certainly exists,
as revealed by the slight birefringence that is often observed for many of these systems.
Our point is to say that, for NMR, these systems must be considered as anisotropic. But
since the medium is generally macroscopicaUy isotropic, an additional 'powder
average' must be performed. Work along these Unes is currently being performed in our
laboratories.

The authors are indebted to Professor R. B. Blumstein and Doctor J. F. d'Allest for
providing the AZA9dl4 polymer sample and Doctor P. Fries for illumina'ang
discussions. This work was partially supported by the Science Program of the
Economic European Community, Contract No. ERBSC1*CTOOS068.

Appendix
To establish equation (24), we use the property that if H1 is a random gaussian
variable with zero average value, the probability density for a given pair of values
-I)=X2 is given by the bidimensional gaussian distribution [16]

i,i-2jwI(l_pl),,2 *i+xj-2px,x21
J'
where
and
The UK- of the relation

I dxl \
J -<D J -a

leads, upon resolution of the double integral, to equation (24) of the text

RefcL
El] DE GE»WES, P. G, 1974. The Physics of liquid Crystals (Clarendon Press).
E2] (=}ÛRSAYljQinDC»YSTALGROup,1969,J.cfcem.P/i>s,5I,816.(b)LEGER,L, 1969,Thesis,
University of Orsay, (c) COLES, H. J., and SEFTON, M. S, 1985,Jtfofec. CrystabUq. Crystals
IfK, 1,151.
£3] (a> DOAME, J. W, 1979, Magnetic Resonance of Phase Transitions, edited by F. J. Owens,
CP.PooleandH.A.Farach(AcademicPress),p.l71.(b)DoNG,R.Y^1983,/s7-.J.C»iein.>
23,370. (c) NOACK, F, 1986, Prog. NMR Spectrosc, 18,171. (d) SCHWHKERT, K. H, and
NOACK, F, 1992, Molec. Crystals liq. Crystals, 212, 33.
[4] WARNER. M, 1984, JIfalec. P/iys, 52,677.
 57

NMR line broadening by elastic modes 665

[5] VOLD. R. U, VOID, R. R, and WAKNER, M, 1988, J. cfcem. Sue. Faraday Trans. 2,94,997
and references therein.
[6] (a) FABER, T. E, 1991, Ltq. Crystals, 9,95 andreferencestherein, (b) References in [5] and
[6(a>] by the same author.
[7] ESNAULT, P., CASQUIUID, J. P.. and VOLINO, F, 1988, Uq. Crystals, 3,142S.
[8] ESNAULT, P, CASQUILHO. J. P, VOUNO, F, and MARTINS, A, F, 1990, Uq. Crystals, 7,607.
[9] (a) CASQUILHO, J. P, ESNAULT, P, VOUNO, F, MAUZAC, M, and RICHARD, H, 1990, Molec.
Crystals Kq. Crystals B, 180, 343. (b) CASQUOHO, J. P, and VOUNO, F., 1990, Molec.
Crystals Uq. Crystals B, 180, 357.
[10] ABRAGAM, A-, 1961. The Principles if Nuclear Magnetism (Clarendon Press), Chap. X.
[11] ABRAMOVTIZ, M, and STEGUN. I, 1968, Handbook of Mathematical Functions (Dover).
[12] GALLAND, D, and VOUNO, F, 1991, J. Phys. U, France, 1,209.
[13] VOLINO, F, and GALLAND, D, 1992, Molec. Crystals £,'. Crystals, 212, 77.
[14] GALLAND, D, GERARD. H, RATIO, J. A, VOLINO, F, and FERREIRA, J. B., 1992,
Afocromofcettles, 25, 4519.
[15] ESNAULT, P., GALLAND, D, VOUNO, F, and BLUMSTEIN, R. B, 1989, Macromolecules, 22,
3734.
[16] KTTTEL, C, 1958, Elementary Statistical Physics (Wiley), p. 139.

\
 CHAPITRE II

CRISTAUX LIQUIDES

•7A.
 *it

6l

g A Blohénvls substitués dans des solvants nematiquesTZll

n A 1 Exploitation des données RMN

Les données de RHN du proton du 4,4'-dichlorobiphényl (DCB) et du
4'-bromo-4-chloro-2,6-difluorobiphényl (DIF), dissouts dans des
solvants nématiques, ont étf analysées en terme du modèle de
conformation unique. Cette étude a été réalisée afin de tester la
validité de ce modèle et de comparer ses résultats à une analyse
précédente basée sur le principe du maximum d'entropie.
Les sept interactions dipolaires proton-proton fournies pour le
DCB ainsi que les douze (dont deux fluor-proton et un fluor-fluor) du
DIF ont été ajustées dans le cadre du modèle à conformation unique en
fonction de cinq et dix paramètres respectivement. Le modèle à
conformation unique donne ici d'excellents résultats.
Il ressort de cette analyse que les structures des deux molécules
sont assez proches, sauf pour les éléments où la présence des deux
atomes de fluor du DIF, de par leur encombrement stérique, agit sur la
géométrie de la molécule, comme on le constate sur les valeurs de la
distance entre les deux phényls ou de l'angle dièdre <)> (~ 33° pour le
DCB et 47.8° pour le DIF). Cette présence se répercute également sur
l'ordre orientâtionnel par le fait que le plan principal XOZ
correspond pratiquement au plan du phényl fluoré.

E A 2 La méthode de maximum d'entropie[22]

Le principe du maximum d'entropie, provenant de la théorie de

l'information, permet de traiter des données expérimentales A(X), où X
est un ensemble de variables, telles que la grandeur A observée soit
la moyenne sur ces variables:

A = <A(X)> = J P(X)A(X) dX (24)

où P(X) est la fonction de distribution normalisée de X.

La méthode de maximum d'entropie va permettre d'obtenir la
distribution Pi-X) compatible avec la donnée A la plus isotrope
s?»
possible. Cette dernière condition est obtenue en maximisant la
fonction d'entropie:

S = J P(X) In [P(X)] dX

Cette distribution PH£ (X) satisfaisant les deux conditions sera

de la forme:

P^(X) =iexp[ - A A ( X ) ]
\
 82

où Z est un facteur de normalisation et A un multiplicateur de

Lagrange déterminé en ajustant la valeur expérimentale A grâce à
l'équation (24). Si Ton dispose de plusieurs données Ak(X), la
fonction de distribution la plus isotrope des variables X pouvant être
déduite des données sera alors:

(25)
TexP

On constats que, par cette méthode, le nombre de paramètres Ak

est toujours éo?l au nombre de données Ak. En l'absence de toute
donnée, la distribution PME (X) sera parfaitement uniforme, et le
résultat de cette méthode est essentiellement une évaluation de la
richesse informationnelle des données traitées vis à vis des variables
utilisées.
Cette méthode de traitement des données a été appliquée aux cas
du DCB et du DIF avec pour unique variable l'angle <t>. Pour le DCB elle
prédit une distribution angulaire centrée en un angle $0 (de l'ordre
de 32°) correspondant à celui provenant du modèle à conformation
unique. Cette concordance incite à penser que la distribution réelle
p(ci>) doit se situer entre la distribution déterminée par la méthode de
maximum d'entropie et le pic delta trouvé par le modèle à conformation
unique (pic élargi par des mouvements librationnels de faible
amplitude). Dans le cas du DIF, la distribution PHE (<t>) est plus large
et présente deux maxima locaux, l'un proche du. 0 trouvé dans notre
analyse (environ 40°), l'autre pour 0°. A partir de simples
considérations stériques cette dernière conformation semble
physiquement peu probable, l'existence de ce maximum ne reflétant que
l'insuffisance d'informations disponibles par la méthode de maximum
d'entropie contenues dans ces douze données RMN au sujet de la
conformation du DIF. Il est probable que si l'on ajoutait Ic condition
que les conformations plates sont négligeables pour le DIF, on
trouverait une distribution PME {")>) de forme semblable à celle trouvée
pour le DCB.

En résumé, la méthode de maximum d'entropie et l'usage de modèles

(dont celui à conformation unique) semblent donc deux voies
J complémentaires pour analyser les données expérimentales, la première
\ pouvant éventuellement supporter un modèle (comme c'est le cas ici,
pour le DCB du moins). Cependant, de par son principe-même, le maximum
d'entropie reste une méthode mathématique mettant en avant la richesse
informationnelle des données traitées et ses résultats doivent être
confrontés à des contraintes supplémentaires (expérimentales ou
théoriques), sans quoi l'on risque d'aboutir à des conclusions peu
plausibles telle l'existence de conformères plats dans le DIF en
solution nématique.

:*:•*
 *£1

63

LIQUID CRYSTALS, 1992, VOL. 12, No. 4,649-656

Hie structure of substituted biphenyls as studied by

proton NMR in nematic solvents
Single conformation model versus
maximum entropy analysis

by H. GERARD, J. AVALOS, D. GALLAND and F. VOLINO*

DRFMC/SESAM/PCM-CENG-85X-38041
F-Grenoble Cedex, France

(Received 24 February 1992; accepted 17 April 1992)

Published [11,12] proton nuclear magnetic resonance data for two substituted
biphenyl molecules dissolved in nematic solvents are analysed in terms of the single
conformation model and compared with the results of the maximum entropy
analysis of [11]. It is shown that (i) this model, in which the number of adjustable
parameters is less than the number of data, can describe very well the data for both
molecules and (ii) the results of the maximum entropy analysis provide global
support for this model. It is argued that the ultimate support of the single
conformation model would be that introduction of a sufficiently large number of
additional data in the maximum entropy analysis leads to a distribution for the
dihedral angle between the two phenyl rings with two symmetrical very sharp peaks.

1. Introduction
The description of non-rigid molecules in fluid phases has attracted considerable
attention for at least two decades. The development of NMR techniques now allows us
to produce accurate data from such molecules in (fluid) anisotropic media [1-3], data
which are highly relevant in this field of research. Surprisingly, there is not yet a
consensus on how these data should be analysed, and disputes and controversies have
plagued the literature for many years. The most typical example is probably that of the
nematic phase of 4-N-pentyl-4'-cyanobiphenyi (SCB) for which a very large amount of
data is available (maybe the largest at present for a small molecule), and for which three
different models have been proposed so far [4-63. Although the various authors
criticize one another, all of these models are based on the rotameric isomeric state
scheme, in which the pentyl chain is assumed to exist in 33 =27 conformations, as in the
isolated molecule. These models correspond to a gas-like picture of a liquid, in which an
isotropic phase is identified to a gas phase, and a fluid anisotropic phase to a biased gas
Phase.
'« In several papers p-lo], we have argued that if such a picture is acceptable to
' describe macroscopic (statistical) properties, it is a very poor description of molecular
properties, and that a solid-like picture in which the molecules exist in essentially one
conformation, as in the solid phase, is preferable. The main arguments to support this
* view lie in the fact that (i) the density ofa liquid is very close to that of the solid existing
at lower temperature and (ii) significant diffuse scattering exists in a liquid at nearly the
same place as that of the Bragg peaks of the solid, suggesting comparable molecular
i> ' 'Author for correspondence.
• ir 0267-S292 J300 © 1992 Taylor & Francis Ltd.

\
•'•f.
 64

650 - H. Gerard et al.

structure and arrangement in both phases. Thus, the local disorder that exists at the
molecular scale in a liquid must be such that it satisfies these two requirements. This has
led to the single conformation model in which the only allowed large amplitude
internal motions (those which produce significant averaging of the magnetic interac-
tions) are symmetry operations such as n-flirs of pheayl rings, methyl group rotation
and exchange with the mirror image conformation (dynamical racemizatlon). Clearly if
additional internal disorder exists in the solid phase, such as end-chain motion, this
disorder survives in the liquid phase. Another consequence of the solid-like picture of a
liquid is that there is no essential difference between isotropic or anisotropic phases at
the molecular scale (the difference is only macroscopic), so that there should not be any
significant contribution of the molecular conformations to the clearing transition
enthalpies and entropies, a prediction which seems to be supported by experiment (a
dimer liquid crystal [9], 5CB and 8CB [5]).
Recently, a different approach has been proposed to analyse NMR data of
substituted biphenyl molecules dissolved in liquid crystals, based on the maximum
entropy principle [U]. From a purely operational viewpoint, the approach may be
considered as a particular many conformation model, in the sense that tl.~ internal
variables which describe the conformations are assumed to be distributed. In molecules
as simple as those considered in [1 1], there is only one internal variable, namely the
dihedral angle, <t>, between the rings. Each value of 4> corresponds to a different
conformation, and the difference between the possible many conformation models lies
in assumptions about the relation between conformation and orientation. In the
maximum entropy analysis, no assumptions of this kind are made, but the least biased
(Le. the flattest and broadest) distribution of the statistical variables (conformational
and orientational) is sought, which is compatible with the data. Gearly, this
distribution cannot be the true distribution since it depends on the nature and number
of the data. However, it is expected that the maximum entropy distribution will tend
towards the true one in the limit of a very large amount of data. Thus the ultimate
support for the single conformation model would be that, in this limit, the maximum
entropy distribution, averaged over the orientational variables, tends towards a (single)
sharp peak.
The purpose of this paper is to test these ideas with two substituted biphenyl
molecules considered in [11].

Z Tkc NMR eata of two Mfastitatedbipbenyl compounds, and analysis in terms of the
single conformation model
2.1. General
We consider the proton NMR data of 4,4'-dichlorobiphenyl (DCB) dissolved in
nematic 152 at three temperatures, and those of 4'-bromo-4-chloro-2,6-
difluorobiphenyl (DIF) in a mixture of nematics, at one temperature. These data are
pubushedin[ll]forDCBandin[12]forDIF.Thestructureofthesetwomoleculesis
sketched in figure 1. Seven different dipole interactions for DCB and twelve different
dipole interactions for DIF have been measured with high accuracy. These data are
reproduced in the table. We show now that these data can be analysed in terms of the
single conformation modeL In order to reduce the number of adjustable parameters, it
is necessary to introduce as much structural information as possible that comes from
sources other than NMR. The most confident source is single crystal X-ray
crystallography.ThestructuresofsolidDCBandsolidDIFarenotknown.Theclosest
 65

Single conformation versus maximum entropy 651

2 1 .5.". 6

6
f

no-
r
Cl
Br.

3 4 8 7
Figure 1. Sketch of the DCB (a) and DIF (b) molecules showing the labelling of the various
spins i and the definition and values of bond distances and angles. The values quoted are
those of biphenyl in the solid phase, taken from [13], and assumed to be fixed in the
calculations.

Dipolar couplings, Dit, as denned in [11], of 4,4-dichIorobiphenyl (DCB) dissolved in the nematic
phase of 152, at three temperatures (rms error. 1-4, 2-5, 5-0 Hz at 312, 322, 332 K,
respectively), and of 4'-bromo-4-chloro-2,6-difluorobiphenyl (DIF) in a mixture of EBBA
and Xl 1643 at 295 K. The observed values are taken from [11] and [12]. The calculated
values are those for the single conformation model discussed in the text

DtfHz

DCB DlF
312K 322K 332 K 295 K

U Obs. CaIc. Obs. CaIc. Obs. CaIc. Obs. CaIc.

1,2 -5374-6 -5374-7 -5167-4 -5167-4 -4922-3 -4922-3 -377O9 -3770-9

U 6»5 691 65« 65-0 596 6Ol 61-2 61-2
si*
Ji
1,4
1,5
1,6
23
477-4 478-1 457-0 456-7 430-5 431-4 360-5
-1908-2 -1908-3 -18207 -182&6 -1720-6 -1720-6 -265-6 -265-3
-446-2 -446-7 -428-3 -429-3 -406-5 -408-7 -264-7 -264-2
479-5 479-1 456« 457-7 432-9 432-3 360-5
360-6

3604
5
W -240-8 -241-7
2,6 -167-9 -164-1 -16»9 -157-8 -154-8 -1503 -113-3 -110-5
5,6 -3157-9 -3157-8
5,7 25-1 27-8
5,8 197-8 197-7 'J
6,7 284-3 283-2
I

if
'1

79
 66

652 H. Gerard et al.

molecule whose structure has been solved by single crystal X-ray crystallography is
biphenyl [13]. This study reveals that the rings are not perfect hexagons, although the
symmetry with respect to the para-axes is preserved. We have fixed the values indicated
in figure 1 to the values found for biphenyl, but left some of the other angles, namely a
and p in DCB and a,, /?„ Ot2, /S2 and yz "> DIF, as adjustable parameters within very
narrow limits. Indeed it is not expected that the mean value of these angles in different
molecules are exactly the same as in biphenyl. Note that these fitted angles are those for
which the différences with biphenyl are presumably most significant (for example near
the fluorine atoms in DIF).
The analysis was performed in the same way as with ethoxybenzene [8], to which
the reader is directed for details. Here the problem is much simpler however. A frame is
attached to one ring in DCB, with Oz along the para-axis and Ox in the plane of the
ring. For DIF, the frame is similarly attached to the fully hydrogenated ring. We have
considered the case of DCB first. This molecule has D4 symmetry, so that the principal
axes of the order tensor are completely determined by symmetry. The principal axis
OZ is the line along the two para-axes, that is Oz1 and OX and O y lie in the two bisector
planes. According to the prescriptions of our model, the only internal motions are
symmetry operations, namely (uncorrelated) it-flips of the two rings around their para-
axes. The exchange with the mirror image is irrelevant here because the corresponding
averaging of the magnetic interactions has already been performed by the it-flips. This
does not mean that such motion is absent, and probably, it actually occurs. The free
parameters of the problem are the two angles a and ft the dihedral angle (j> between the
two rings and the two components Tlx and T1x- Tn of the order tensor. Because of
the D4 symmetry, T1,=1(7^-7;,)tan(^), T^=T1x=O. In total, there are five free
parameters to be determined by seven pieces of data, a situation that is similar to that of
ethoxybenzene [8]. In fact, the test is more severe here since the data exist for three
temperatures, and only the two order tensor components, and the dihedral angle <f>
are allowed to vary.
The case of DIF was also examined. The symmetry is now only D2. The principal
axis OZ is still along the line along the two para-axes, implying T11—T1x=O as for DCB1
but the three other components now need to be considered as adjustable parameters. In
addition to the five structural angles mentioned previously, the two other geometrical
parameters are the dihedral angle <j> (as for DCB), and the value of the CC distance
between the two rings J00 which could not befixedto the value of biphenyl (1-495 A). In
total, there are ten parameters to be determined by twelve pieces of data, again a
situation similar to that of ethoxybenzene [S].

2.2. Results
22.1. DCB
The best fitted values of the geometrical parameters are «= 119-7°, ft= 119-5° and
#=32-51°, 32-62°, 32-72° at 312,322,332 K, respectively. The two components of the
order tensor T11 and T1x-Tn are 0-6629 and 00416; 0-6374 and 0-04381; f>6071 and
0-04656 at 312, 322 and 332K, respectively. The values of the dipole interactions
calculated with these values of the parameters are given in the table.
These results show that the molecular properties, namely the structure, but also the
conformation are independent of temperature. The overall variation of $ of 0-2° over
20 K. can indeed be considered as negligible since fits of practically the same quality are
obtained by fixing this angle to the average value. In contrast, the statistical properties

f
 67

Singie conformation versus maximum entropy 653

of the nematic phase pictured by the values of the order tensor components are
strong functions of temperature.
Diagonalization of the order tensor yields the principal frame OXYZ and the
uniaxial and biaxial order parameters Szz and Sxx—S^y. As already mentioned, the
principal planes XOZ and YOZ are the two bisector planes of the rings. With the
convention that OX and O Y are such that Sxx—Sn. is positive (OX is less ordered than
OY),ZOX is the bisector plane corresponding roughly to the plane of the molecule. The
uniaxial and biaxial order parameters are: 0-6629 and 0-0489; 0-6374 and 0-0526; 0-6071
and 00554 at 312,322 and 332 K, respectively (note that Szz=T11 for evident reasons).
It is observed that Szz decreases and that Sxx-Sït increases slightly with increasing
temperature. This result is in agreement with theory [14-161, and observed in
other nematic phases for non-rigid molecules [7,17] (this behaviour is common for
rigid molecules, see, for example, [IS]).

222. DlF
The best fitted values of the structural parameters are «1 = 119-6°, />, = 119-5°,
a2= 119-1°, P1= 121-7°, y2=120-7°, rfcc= 1-512 A. The dihedral angle 4> is 47-80°. The
values of the three order tensor components T11, T1x-Tn and Tx, are 0-464, -0-004
and 04)40, respectively. The calculated values of the dipole interactions are given in the
table. Diagonalization of this tensor yields the principal axes and the two order
parameters. The principal plane XOZ is between the two rings as in DCB, but now very
close to the plane of the fluorinated ring (at 1-29° from the fluorinated ring and 46-51°
from the protonated ring: the sum is just 47-80°). The order parameters Szz and Sxx
-Sn are 0-464 and OO80, respectively.

2.3. Discussion
These results show that the single conformation model is sufficient to describe
satisfactorily the sets of NMR data for the two molecules considered. Indeed nothing
unexpected is predicted by this model.
The values found for the angles ot, f, y are comparable to those of biphenyl, and can
thus be considered as acceptable in the absence of further structural information. The
fact that the distance 4cc in DIF is found to be slightly larger than in DCB and in
biphenyl is consistent with the presence of the two bulky fluorine atoms. Thus, the
structural angles and distances are very similar in all three molecules. In contrast, the
conformations are different. This is not surprising since crystallographic data of other
substituted biphenyls show that the angle 4> can lie between about 10° in biphenyl [13]
to values larger than 40° in cyanobiphenyl compounds [19-21], (the value for biphenyl
in the gas phase is close to 45° [22]), emphasizing the importance of packing forces to
determine this angle in condensed matter.
Concerning the orientational order, the most interesting result is probably the
if»
finding that the principal plane JfOZ in DIF is practically the plane of the fluorinated
•I ring. All of these results appear as reasonable support for the single conformation
model for these molecules. The quality of the test is now checked with the results of the
maximum entropy analysis.

3. Comparison with the maximum entropy results

In figure 2 are reproduced the distribution functions p(tj>) predicted by the analysis
of[l I], and the distribution predicted by the single conformation model, namely two
delta peaks at # and #+jc, for DCB and DIF.

1
 68

654 H. Gerard et al.

30 60 90 120 150 180

0.6

0.4 -

0.2

Figure 2. Probability distributions of the dihedral angle # between the two phenyl rings of
DCB and DIF, predicted by the maximum entropy treatment (reproduced from [H]).
Note that for DCB, the width of the distribution increases as the temperature increases, i.e.
as the order, picturing the amount of information, decreases (as expected). The vertical
lines correspond to the single conformation mode) (delta peaks) and the horizontal dashed
line (ordinale l/2>r=0-159...) corresponds to the uniform distribution (maximum entropy
distribution in the absence of any data).

For DCB, the maximum entropy analysis appears to provide good support for the
singje conformation model, in the sense that it seems to exclude any conformation
model with two or several conformations with very different values of the dihedral
angle. The fact that the delta functions correspond to the two maxima of the
distribution and are practically independent of temperature, suggests that the shape
(not the height!) of the distribution is probably close to that of the true distribution (the
peaks are broadened by small amplitude librational motions).
For DIF, the situation is different. The distribution P(IJI) is broader, with two
maxima. The main maximum is dose, but not identical, to the position of the delta
function, and the secondary maximum is at O (or it). Existence of these two maxima may
mean that (i) this molecule exists in essentially two conformations, and if so, (ii) the fact

\
 69

Single conformation versus maximum entropy 655

that the single conformation model works well, is fortuitous. However, is is clear from
steric considerations that the internal energy for a flat conformation is large in DIF
[12], implying that the probability is very weak for ^=O. Thus, this secondary maximum
probably does not reflect any physics, but particular features of the NMR data. This
result probably also means that the amount of molecular information contained in the
twelve pieces of data for DIF is less than the seven data for DCB, a situation which
seems to be paradoxical. The contradiction may, however, be only apparent, because
the symmetry is lower in DIF.
At this stage, it is worth noting that the maximum entropy analyses of [11] has been
performed with slightly different geometries, so that, strictly speaking, the comparison
cannot be made. However, because the theoretical values of the inter-ring interactions,
which contain the information on p(<W, are little affected by small changes in the angles
and distances, it can reasonably be expected that the maximum entropy analysis
repeated with the present geometries will yield essentially the same results.
The result for DIF emphasizes the fact that the distribution does not correspond to
a physical quantity, but to our degree of knowledge about the system after analysis of
the corresponding data. If this degree is large, as seems to be the case for DCB, the
maximum entropy distribution may be close to the true distribution. If the degree is
small, as for DIF, the similarity is only rough, and incorrect conclusions may be drawn
(for example the existence of flat conformers in DIF) if the basic principles of the
maximum entropy analysis are not kept in mind. In this context, it would be very
interesting to see how the distribution changes, if the additional piece of data,
constituted by the condition that the probability of occurrence of flat conformations in
very small, is introduced in the analysis.
In conclusion, it appears that modelling and the maximum entropy method are two
complementary ways to analyse experimental (NMR or other) data. The latter method
may possibly be a way to support some models (as is the case with the single
conformation model for DCB and to a lesser extent, for DIF). or reject others.
Nonetheless, the simplest manner to test models is to study their predictive character,
by including more and more independent pieces of data into the analysis without
increasing the number of parameters. Combination of both methods is certainly the
most convincing way to establish the validity of models.

Fruitful discussions with Drs A. J. Dianoux and C. M. E. Zeyen are acknowledged.

Correspondence with Professor C Zannoni has been highly appreciated. This work
was partially supported by NSF/CNRS Grant INT-871501/88-6920156.

References
[I] EMSLEV, J. W, and LINDON, J. C, 1975, NMR Spectroscopy Using Uquid Crystal Solvents
(Pergamon).
[2] DIEHL, P, and KHETRAPAL, C. L, 1969, NMR Basic Principles and Progress, edited by P.
Diehl, E. Fluck and R. Kosfeld (Springer-Verlag).
[3] DOANE, J. W_ 1979, Magnetic Resonance of Phase Transitions, edited by F. J Owens C P
Poole, Jr. and H. A. Farach (Academic Press), p. 171.
4-: [4] SAHULSKI, E. T, and DONG, R. Y, 1982, J. diem. Pfcju, 77,5090.
[5] CbUNSEU, C. J. R, EMSLEY, J. W., HEATON, N. J, and LUCKHURST, G. R., 1985, Molec.
Phys^ 54, 847.
[6] ABE, A^ and FURUYA, H, 1988. Molec. Crystals liq. Crystals, 159, 99.
LTJ GALLAND, D-, and VOLINO, F., 1989, J. Phys^ France, 50, 1743.
[8j GALLAND, D, and VOLINO, F, 1991, J. Phys. II, France, 1,209.

\
 '•f.

70

656 Single conformation versus maximum entropy

[9] VOLINO, R, RATIO, J. A, GALLAND, D, ESNAULT, P., and VOLINO, F., 1990, Molec Crystals
Kq. Crystals, 191,123.
[10] VOLINO, F, 1991, Proceedings of the European Conference on Liquid Crystals, March,
Courmayeur, Italy, 1992, Molec. Crystals liq. Crystals, 212, 77.
[II] CATALANO, D., DI BARI, L, VERACINI, C A., SHILSTONE, G. K, and ZANNONI, C, 1991,
J.chem.Phys.,94,192%.
[12] FIELD, L. D., and STERNHELL, S, 1981, J. Am. chem. Soc., 103, 738.
[13] CAILLEAU, H., BAUDOUK, I. L, and ZEVEN, C M. E, 1979, Acta crystallogr. B, 35,426.
[14] STRALEV, J. P., 1974, Phys. Rev. A, 10,1881.
[15] LUCKHURST,G.R,ZANNONI,C,NORDIO,P.J.,andSEGRE,U., 197S,Molec.Phys.,30,1345.
[16] BERGENSEN, B., PALFFV-MUHORAY, P, and DUNMUR, D. A., 1985, Molec. Crystals liq.
Crystals, 129,375.
[17] BAD-GANG Wu, ZIEHNICKA, B, and DOANE, J. W, 1988, J. chem. Phys^ 88,1373.
[18] EMSLEY, J. W, HASHIM, R., LUCKHURST, G. R, RUMBLES, G. N, and VILORIA, F. R, 1983,
Molec. Pfcjis, 49,1321.
[19] VANI. G. V, 1983, Molec. Crystals liq. Crystals, 99,21.
[20] HAASE, W., PAULUS, H, and PENDZIALEK, R, 1983. Mofcc. Crystals liq. Crystals, 100,211.
[21] TASHIRO, K, JIAN-AN Hou, KOBAYASHI, M., and INOUE, T, 1990, J. Am. chem. Soc., 112,
8273.
[22] ALMENNINGEN, A, BASTUNSEN, O, FERNHOLT, U CYVIN, B. N, CYVIN, S. J, and SAMDAL, S.,
1985, J. molec. Struct, 128, 59.

V
t
 71

II B Pentvl-Cvanoblphényir231

H B l Analyse des données RHN

Le 4-n-pentyl-4'-cyanobiphényl (5CB) est certainement Tune des

molécules, présentant une phase nématique aux alentours de l'ambiante,
pour laquelle un maximum de données RMN (et autres) est disponible.
Cette abondance de données (proton, deuterium et carbone 13) doit
permettre d'effectuer un test sévère du modèle à conformation unique,
puisqu'à une température ( T01 - 5K) on dispose de 32 données
différentes (mais non forcément indépendantes) à analyser, dans le
cadre du modèle, avec moins de vingt paramètres (les cinq composantes
du tenseur d'ordre ainsi que divers paramètres géométriques dont
l'angle dièdre <f> entre les plans des deux phényls), ainsi que de 19
données (proton et deuterium) pour une gamme de sept autres
1
températures .
' Le résultat principal de cette étude réside en l'échec apparent
du modèle à conformation unique: on ne peut reproduire les données RMN
du 5CB avec une seule conformation.
Cette constatation devait mener soit à abandonner ce modèle qui
s'était pourtant déjà révélé utile dans le cas de plusieurs molécules,
de complexité certes moindre que celle du 5CB, soit à l'améliorer
d'une quelconque façon. Pour cela, il faut revenir au principe même du
modèle, qui est qu'en phase nématique les molécules doivent avoir une
conformation proche de celle présentée en phase solide. Dans les
molécules étudiées jusqu'ici, la conformation en phase solide révèle
une conformation "unique" (double pour les molécules chirales, avec la
conformation racémique), mais il est des cas où des molécules,
nématogènes ou non, présentent un désordre de bout de chai ne dans
cette phase, tel le terephtal-bis-butylaniline (TBBA) ou le
butyl -cyanobiphényl (4CB). Ce désordre, qui correspond en pratique à
une délocalisation du carbone methyl i que se trouvant en fin de chaîne,
a été mis en évidence par diffusion quasi -élastique de neutrons et
diffraction de rayons X sur des monocristaux. Suivant la philosophie
îj* du modèle si un tel désordre existe en phase solide, il doit persister
| à plus haute température (et éventuellement s'amplifier) dans les
. phases fluides. La présence d'un tel désordre dans le 4CB nous a donc
- ; incité à supposer son existence dans la phase solide, et a fortiori
nématique, du 5CB, désordre que nous avons choisi de circonscrire aux
deux carbones terminaux.
Le modèle à deux sites utilisé (inspiré de l'analyse des données
\- cri stall ographiques du 4CB) a permis de décrire les données du 5CB de
;jr manière satisfaisante.
La conformation trouvée pour le groupement biphényl, d'abord,
n'apporte pas en soi d'éléments très nouveaux. L'angle <f de 32.21° est

¥
 72

comparable aux angles dièdres précédemment trouvés pour d'autres

molécules à base de biphényls, et le fait qu'on puisse ajuster les
données en fonction de la température sans le faire varier avait déjà
été pressenti lors de l'analyse des données de biphényls substitués.
Plus intéressants sont les résultats concernant la chaîne
aliphatique. Le résultat le plus prometteur de cette étude reste
évidemment la prédiction d'un désordre de bout de chaîne, sans lequel
les données ne peuvent être décrites par un modèle à conformation
unique (au sens littéral du terme). Ce modèle de désordre donne des
résultats très similaires quel que soit le sous-modèle employé et
qu'on peut comparer à ceux du 4CB.
Les deux positions déterminées correspondent respectivement à un
angle de torsion proche de 0° (défini par rapport à la position trans)
associé à un angle de valence CCC de valeur bien inférieure à celle
généralement observée dans les chaînes alkyls des nCB et à un angle de
torsion plus élevé associé à un angle de valence de valeur standard.
Ceci est aussi le cas dans le 4CB pour lequel la conformation "trans"
est la plus probable, comme dans le cas du 5CB où celle-ci, de l'ordre
de 0.75, décroît quand la température augmente.
Enfin nous avons obtenu le tenseur d'ordre complet sur toute la
gamme de températures de la phase nématique. Chacune de ses composante
décroît (ou du moins ne croît pas), en accord avec une diminution
générale de l'ordre au sein de la phase nématique à mesure que l'on
approche de la transition nématique-isotrope. D'autre part
l'orientation du référentiel principal du tenseur d'ordre par rapport
au référentiel moléculaire lié au phényl comprenant le groupe cyano
varie, quoique légèrement, en fonction de la température.
 jsu* JErL ...

73

THE MOLECULAR PHYSICS OF NEMATIC PENTYL-CYANOBIPHENYL (5CB)

REVISITED IN TERMS OF THE SINGLE CONFORMATION MODEL:
!-CONFORMATION AND ORDER TENSOR FROM ANALYSIS OF NMR DATA

H.Gérard, D.Gall and and F.VoIino"

DRFMC/SESAM/PCM - CENG - 85X - 38041 F-Grencble Cedex

ABSTRACT

The complete sets of NMR data of Emsley et al [10], Sinton et al [11]

and Fung et al [12] about nematic 5CB, and involving fourty different
magnetic interactions, are tentatively analyzed in terms of the single
conformation model. However, not any satisfactory fit can be obtained in
the strict framework of this model. We show that the reason for this
failure in the case of 5CB is probably due to the existence of some
end-chain disorder. The analysis was thus performed with an approximate
version of the single conformation model in which this disorder is
described inside a molecular frame whose orientational order is still
described by a single order tensor. In this way, the conformation of the
molecule, the parameters describing the end-chain disorder, and the
complete order tensor could be determined over the whole nematic range.
The dihedral angle between the rings could be kept fixed. The torsional
angles in the pentyl chain are found to (slightly) change in such a way
tbit the internal molecular energy decreases on heating. Diagonalization
of the order tensor shows that not only the two order parameters, but
also the directions of the principal axes change with temperature.
Comparison with results obtained with three different "many
conformation" models, and with predictions of these models concerning
the clearing enthalpy and entropy is also made.
The most interesting result of this study probably stands in the
reason invoked to explain the failure of the model for the particular
case of 5CB, namely the existence of an end-chain disorder, and in the
fact that this disorder appears to be very similar to that reported in
the solid phase of the parent molecule 4CB. If this disorder is

'I
i**
confirmed in 5CB, this would show that, due to the very constraining
conditions of its applicability, the single conformation model could
constitute an excellent means to put in evidence peculiar dynamical
phenomena.

Member of CNRS, France.

i
Short title: MOLECULAR PHYSICS OF 5CB REVISITED: NMR DATA

t Classification Physics Abstracts: 61.16N 61.30

 SfL

74

1. INTRODUCTION

The description of nonrigid molecules in condensed fluid media is a

complex and still controversial subject. In several papers [1-6], wehave
shown that the NMR data of a number of nonrigid molecules in liquid
crystalline phases can be described assuming that the only large
amplitude internal motions are symmetry operations. In intrinsically
nonchiral molecules, these operations are mainly ir-flips of phenyl
rings, methyl group rotations and exchange with the mirror image
conformation. This last motion - the parity operation- is conveniently
called racemization motion. It is not claimed that small amplitude
librational motions do not exist, they are simply neglected in the model
because their averaging effect on magnetic interactions is small. A
model with these features has been called single conformation model, to
emphasize the fact that knowledge of one conformation (say the dextro
conformation) is sufficient to define the laevo conformation. A severe
test of this model has been made with molecules based on the
ethoxybenzene moiety [3] and on the biphenyl moiety [6] in situations
where the number of independent data is larger than the number of
adjustable parameters.
In the papers mentioned, and especially in [4], arguments have been
given to support the idea that this model is not only operational, but
corresponds to some physical reality. These arguments are essentially
based on the fact that the .molecular conformations and local arrangement
in a liquid phase, whether isotropic or anisotropic, are (very) similar
to those existing in the solid phase at lower temperature (solid-like
picture of a liquid).
In this paper, we wish to explore further this model by testing it on
the molecule for which probably the largest number of NMR (and other)
data exists so far, namely nematic pentyl-cyanobiphenyl (5CB). We show
that this model can describe the whole set of NMR data, except those
associated with the last méthylène and methyl groups of the pentyl
chain. Agreement is achieved only if dynamical disorder of the methyl
carbon is assumed. Despite the fact that this result constitutes a
counter-example to the single conformation model in its strict sense,
the reason for its non-validity is clearly due to a specific phenomenon
of the associated solid phase, and in this sense, this failure may be
considered as a prediction of the model. It is indeed remarkable that
this prediction has been observed in solid phases of some mesogenic
compounds.
In Section 2, the 5CB molecule is described. The relevant NHR data are
listed in Section 3 and the basic formulae are given in Section 4. The
data are analyzed in Section 5 and discussed in Section 6. Comparison
with the results of analyses with three "many conformation" models is
made in Section 7 and concluding remarks are given in Section 8.
 75

2. THE 5CB MOLECULE

Figure Ia,b shows a sketch of the 5CB molecule and defines the
labelling of the various spins. The carbon spin associated with a proton
(or deuteron) will have the same label "prime". The molecule is composed
of two moieties, the cyano-biphenyl moiety and the pentyl chain. The
bond angles and distances mentioned in the figure have been chosen
according to the ranges of values deduced from single crystal
diffraction data of similar compounds [7-9], and fixed to the particular
value quoted, in order to achieve the best fit. Slight changes in these
values do not however affect significantly the results of this study. A
cartesian frame Oxyz is attached to phenyl ring A (close to the cyano
group), with Oz along the para-axis and Ox in the ring plane. The
geometry is assumed to be such that the first CC bond of the chain is
colinear with the common para-axis of rings A and B. This symmetry is
suggested by the crystallographic data of 3CB [8] and 4CB [9], which
shows that in the solid phase of these compounds, distortions at these
levels are weak.

FIGURE 1

The geometrical parameters of the problem to be determined by the fit

are the torsional angles which define the "single conformation", namely
the dihedral angle <J> between the two rings, and the torsional angles <p^
A = a to w, for the pentyl chain (the convention is (P01= O for the two
first CC bonds of the chain lying in the plane of ring B, and ^= O (A =
3 to ça) for the all-trans chain). The sign of these angles are such that
the successive rotation axes (the para-axes of the rings and the CC
bonds of the chain) are oriented from the cyano group towards the
extremity of the chain (left to right in figure 1). The dynamical
(statistical) parameters are the five elements, in the Oxyz frame, of
the single tensor order. Contrarily to the bond angles and distances,
these latter quantities are expected to be temperature dependent. This
dependence is presumably weak for the torsional angles, and such that it
corresponds to a decrease of the intramolecular energy as temperature
*j* increases. For the order tensor on the contrary, this dependence may be
I rather strong, particularly near the clearing point.
Vi
-' 3. THE DATA

The following published data sets will be analyzed.

.-,' Set (i): the proton NMR data of 5CBdIS (deuterated everywhere except on
* ring A), obtained under the condition of deuterium decoupling [10], and
*, reproduced in figure 2.

8
 76

Set (ii): the deuterium NMR data of 5CBdI5 ( ring B and pentyl chain)
obtained under usual operating conditions [10], and reproduced in
figures 3, 4 and 5.
Set (iii): the deuterium NMR data of BCBdS (deuterated on the two rings;
usual operating conditions [1O]. They are reproduced in figures 5 and 6.
Set (iv): the multi-quanta proton NMF data of 5CBdIl (deuterated on the
pentyl chain only), taken at c.a. 13 K below TJIl]. They are reproduced
in Table 1.
Set (v): the 13 C carbon NMR data of fully protonated 5CB, obtained under
the conditions of removal homonuclear dipolar coupling and use of
off-magic angle spinning, yielding the ortho and vicinal 13 C proton
dipolar interactions on the two phenyl rings, at c.a. 4 K below Tc [12].
They are reproduced in Table 2.
Set (vi): same as set (v) for the pentyl chain [12], also in Table 2.

4. BASIC EQUATIONS

The two basic equations that are useful concern the average values of
the dipolar and quadrupolar interactions. The dipolar interaction D1 .
between spins i and j, and the quadrupolar interaction acting on
deuteron i, are given by (see e.g. [3, 13, 14]):
r, P2(COSB) l sin2B cos2A

sin2B sin2A _ sin2B sinA _ sin2B cosA , ,

< 3
> T y+ < 3
> Tyz + < >T ] (1)
r. * r. r3.
and
I
Q-i = ci [<P2(cosB)> T22+ ^.<sin2B cos2A> (Txx- Tyy) +

<sin2B sin2A> T xy+ <sin2B sinA> Ty2+ <sin2B cosA> T2x] (2)

In these equations, Tr1 ^ are the gyromagnetic ratios of spins i,j

(subscripts P, D or C when referring more specifically to proton,
4* deuteron or 13C spins) and B, A are the polar and azimuthal angles of
•'l vector ij (Eq. 1) or of CD bond i (Eq. 2} in the Oxyz frame; r^ is the
' distance between spins i and j and Tap (a,p = x,y,z are the components
of the order
Vl tensor T of the "single conformation" in the Oxyz frame.
--' The brackets stand for an average over the internal motions. Since,
according to our model, these latter motions are symmetry operations,
namely (independent) ir-flips of the two rings, methyl group rotation and
racemization motion, this average is simply made by averaging the
corresponding quantities over the spins which exchange under these
operations. The other symbols have their usual meaning [3]. The
numerical values relevant to our problem are h-r2,/ 4^2 = 120.12,
 77

h-Yp-Yc/ 4-n2 = 30.211 and h-vg/ 4^2 = Z.832, all in kHz.A3UnHs, and C1 =
172 or 185 kHz for aliphatic or aromatic deuterons respectively [13,14].
Note that these equations are strictly valid only if the tensors have
cylindrical symmetry around the vectors mentioned. This is rigorous for
the dipolar tensor but only approximate for the deuteron quadrupolar
tensors. This aspect is discussed in Section 5.4 in connection with the
rings quadrupolar data.

5. ANALYSIS

5.1 GENERAL

The basic assumptions that will be made is that partial or tctal

deuteration do not change (i) the conformation of the molecule and (ii)
the statistical properties of the phase. The latter assumption means
that the dependence of the order tensor with reduced temperature
t = T - Tc is the same for all compounds.
A problem which arises with data coming from different sources is the
temperature scaling. In NMR spectrometers, it is generally very
difficult to know the actual sample temperature to an accuracy better
than ~ 1 or 2 K, especially when extra-heating is present due to the
strong radiofrequency field of spin decoupling (case of data sets (i),
(iv),(v), (vi)). Thus, the nominal values of the reduced temperatures
quoted in the corresponding references cannot be taken too literally. In
fact, the more correct values are probably for those sets of data
obtained without such extra-heating, and we have assumed that the
"correct" temperature scale is the one of the deuterium NMR experiment
corresponding to sets (ii) and /or (iii). In the analysis, the nominal
values of the other sets have been allowed to be off by no more than two
K with respect to this reference scale.

5.1 DIPOLAR DATA ASSOCIATED WITH THE BIPHENYL MOIETY

The dipolar data of sets (i), (iv) and (v) are considered.
Set (i) gives three proton-proton intra dipolar interactions for ring A
1
1
at various temperatures. Set (iv) gives three proton-proton dipolar
interactions for ring A, three similar interactions for ring B and three
inter-ring dipolar interactions at t ~ - 13 K. Finally, set (v) gives
four 13C -proton dipolar interactions for ring A and four similar
interactions for ring B . Two small problems arise with the former data
[U].
The first problem is that the values of the interactions deduced from
"the multi-quanta spectra depend on the assumption made concerning the
symmetry of the biphenyl moiety. Two symmetries have been assumed,
namely the D4 and the D7 symmetries. A priori, neither of these two

If
 78

f
symmetries is correct since the 5CB molecule having no symmetry at all,
there is no reason for the order tensor to have its main principal axis
02 colinear with the common para-axis Oz, as implied by these
symmetries. However, because of the two indépendant ring TT-flips, the
coefficients of Tyz and T2x in Eqs.(l) and (2) are zero for these
interactions, and only the first three terms survive. This situation is
mathematically equivalent to an effective D2 symmetry for the spin
hamiltonian associated with biphenyl moiety, and this justifies a
posteriori the assumed Q2 symmetry. Concerning the possible higher
symmetry D4, it is not justified by the 13C and deuterium data (sets (v)
and (Ui)), which show that the two rings are not equivalent. However,
this non-equivalence shows up very weakly in the protun spectra [11],
and from this viewpoint, the D4 symmetry can be considered as a
reasonnable approximation.
The second problem concerns the values of two interactions D14 and D23
between ineta protons on ring A, the same two interactions D58 and D67 on
ring B, and the intermediate inter-ring interactions D16 and D25. The
actual values of ther.e i:.teractions imply unreasonably large distortions
of the two rings. In TO-"-, as suggested in [11], tli?> explanation for
this situation is that Lhe multi-quanta spectra depend main'iy oi. the
average value of these interactions and very little on the individual
values. We have verified that this property holds for the normal
(single-quantum) spectrum. This result is also consistent with the fact
that it seems impossible, to measure separately D14 and D23 in the four
proton spectrum [10] . Thus the values of the interactions mentioned
should be replaced by their average value, in the three cases
considered. A support for this procedure is that the average >alue
- (D14+ D 23) is now completely consistent with the corresponding value
of set (i).

The actual analysis is made as follows: the phenyl rings are assumed
to be perfect hexagons with a CC distance of 1.40 A, the slight
distortions being introduced in the values of the inclinations u|T (i =
1, 2, 5, 6) of the CH bonds on the para-axis. These u!T are taken as
(temperature independent) adjustable parameters within very narrow
ranges (± 1°) around 60°. The two other constant parameters are the
dihedral angle $ between the two rings and the distance dcc between the
*i two rings. The temperature dependent parameters are the three order
' tensor elements T2z, Txx- Tyy and Txy(as explained above, the two other
%; elements do not appear in the expressions of interactions associated
^ with the biphenyl moiety). The three sets of data were considered
simultaneoulsly in the fitting procedure, allowing the temperatures of
the various sets to be off by no more than 2 K with respect to their
• nominal values. The output of the calculation are the values of the
A parameters mentioned.
v
i Satisfactory fits are obtained, despite the fact that the number of
*, free parameters is less than the number of data (a similar situation was

92
 79

obtained with other molecules based on the biphenyl moiety [6]. The
constant parameters are u\ = 59.06°, U^ = 60.04°, U^ = 59.79°, u£ =
59.77°, <{> = - 32.21°, dcc = 1.49 A, and the tensor elements are given in
figure 9. The uncertainty on the latter are estimated by repeating the
fitting procedure taking different values of the measured interactions
within the experimental error bar.
The quality of the fits can be appreciated in Tables 1 and 2 and in
Figure 2, where are compared experimental and calculated values of the
interactions, corresponding to set sets (i), (v) and (vi), respectively.
In Table 1 are mentioned the values of the interactions obtained from
the multi-quanta spectra under both assumptions of D2 and D4 symmetries,
in order to estimate the magnitude of experimental systematic errors. It
is seen that observed and calculated values are in agreement with one
another within this error bar.

FIGURE 2

TABLE 1

TABLE 2

Before ending this section, it is worth noting that the choice of the
sign of <f could have been opposite. But once chosen, this sign should
not be changed anymore since it fixes the signs of the order tensor
elements and the signs of the <p^ determined below. The conformation is
defined by the set of angles {<f, <PX }• The conformation defined
by {- <t>, - ^) is just the mirror image conformation. Both conformations
are equivalent and exchange fast on the NMR time scale.

5.2 DIPOLAR AND QUADRUPOLAR DATA ASSOCIATED WITH THE PENTYL CHAIN.
The
\-\ data sets (ii) (pentyl chain data only) and (vi) are considered. "
J Set (ii) gives the quadrupolar interactions qx and the DD dipolar VJ
interactions Dx (A = a to u) associated with the méthylène and methyl -'
deuterons of the fully deuterated pentyl chain, at several temperatures.
These temperatures will be considered as the "correct" ones, as defined
Y above. More specifically, eigth values of reduced temperatures between O \-
and
'h " 17 K are considered in practice. Set (vi) gives the 13C-proton V
» interactions Dx.x within the méthylène and the methyl groups, atone '\
temperature ~ - 4 K. I

93
 80

Contrarily to the biphenyl data, there is no special problem with

these data. The only problem that appears when attempting to describe
simultaneously all these data in terms of the model, taking the four
torsional angles (pa (a = 1 to 4), and the remaining order tensor
components as adjustable parameters, is that this is completely
impossible! There is no way to approach (even very roughly) the actual
values of the interactions, whatever the values chosen for the
structural angles CCC or HCH of the chain. This clearly is an indication
of the break-down of the single conformation model, at least in its
strict sense. Introduction of librational motions around the mean values
of the torsional angles does not improve the situation, incidently
showing the little averaging effect of this kind of motions.
At this point, either the model should be abandonned, or modified in
some way. A hint concerning a possible modification was the following.
The basic philosophy underlying the single conformation model is that
there is one conformation in the liquid phase because there is one
conformation in the solid phase at lower temperature [4]. This may be
the situation in 5CB since there is evidence of end-chain disorder in
some solids exhibiting mesophases at higher temperature. The two most
striking examples are terephtal-bis-butylam'line (TBBA) [13,14], and
butyl-cyanobiphenyl (4CB) [9]. This evidence has been obtained by
quasi-elastic neutron scattering [13] and single crystal X-ray
diffraction [9,14]. If such disorder, corresponding in practice to a
délocalisation of the methyl carbon, exists in the solid, it must also
exist in the liquid. The consequence of this situation for 5CB is that
the data associated with this possible (extra) disorder, namely those
attached to carbon atoms S and u, should in a first stage, be excluded
from the single conformation analysis.
First Stage.

In this first stage, the data associated with the three first carbons
a, p, TT are considered to determine the three torsional angles <pa , cpp
and <pT , and the matrix elements Tyz, Tzx, and Txy (Note that the values
of this last component were estimated at t — 13 and - 4 K using sets
(iv) and (v). The values of the tensor elements T22, and T xx - T are
fixed to values previously obtained. Again, the temperatures are allowed
i*
to be off by no more than two K between the nominal values of the
'i various sets. This will produce the final renormalized temperatures for
the sets (i), (iv) and (v). All the figures in the paper are drawn using
this renormalized temperature scale.
Good fits are obtained and all the parameters could be estimated. The
structural angles Ux = CCC and HCH are chosen within the ranges
permitted by the literature, namely ua = 110°, UB = UT = 114.5°, HCH =
108.5°. These values are those which produce the best fits. Contrarily
to the dihedral angle between the rings, the torsional angles are found
1. to be slightly temperature dependent. Evidently the order tensor
elements are also found to be temperature dependent. The results are
 81

shown in figures 7 and 9. The quality of the fit can be appreciated in

Table 2 and in figures 3 and 4, where are compared experimental and
calculated values. Concerning the temperature scales, the nominal values
of set (i) appear to be systematically too large (the sample is hotter
than quoted in [1O]). The difference is very small close to the clearing
point, but it increases and reaches about 2 K at t ~ - 18 K. Set (iv)
[11] corresponds to a reduced temperature of -11 rather than -13 K, thus
reducing the discrepancy between sets (iv) and (i). Finally, set (vi)
corresponds to — 5 rather than to - 4 K [12].

FIGURE 3

FIGURE 4

Second stage

In this second stage, an attempt is made to analyze the remaining data

associated with the last méthylène and methyl groups. Here, the
interactions are further averaged by an additional internal motion. The
difference between this latter motion and the internal motions allowed
by the single conformation model is that it is no more a symmetry
operation, at least at the scale of one molecule. The simplest way to
describe such motion is to assume that the torsional angle <ps can take
two values <pg0 and Cp51 with probabilities p and 1 - p. However, this
more complex description is not sufficient to supply a satisfactory
solution, whatever the value chosen for the angle U5. A model where <py
is uniformely distributed over 2tr can also be excluded.
The hint came from the crystallographic data of 4CB [9]. The analysis
of these data made in [9] indeed suggests that the end-chain disorder
should be analyzed in terms of the above two site model, but
quantitative agreement only be obtained assuming two different values
for the angle U5. We thus similarly introduced two values U50 and u si in
our model, but in this case, an infinite number of solutions with two
conformations are possible. We have explored two of them, labelled I
and II, defined fay the additional constraint: Cp51 = (P80+ -rr and Cp80 = O ,
respectively. Calculation of the interactions averaged by exchange
between the two conformations has been made using the single order
tensor determined previously. It is clear that two order tensor would
have been necessary to descride completely the orientational order of
each of these conformations, the tensor optimised for the rest of the

\ molecule being the average of these two tensors. The succès of the fit
in the framework of this average tensor simply shows that the two
tensors differ slightly, or at least, that the data considered do not
 82

allow to discriminate one from the other.

For the two sub-models I and II, very good fits are obtained. The
three angles as well as the probability p are found to be weakly
temperature dependent as shown in figures 7 and 8a,b. The experimental
and calculated values of the interactions are reproduced in figures 3
and 4. As above, the uncertainty on the values of the various parameters
are estimated by repeating the fits with different values of the
measured interactions, and of the order tensor elements, within the
error bars.
It is worth noting that these two sub-models chosen are not unique to
describe the end-chain disorder. As stated above, their choice has been
suggested by the results on 4CB [9]. However, a model such as that
suggested by TBBA [14] would probably work as well. At this level of
modelling, use of such more complex model is not justified.

5.4 QUADRUPOLAR AND DIPOLAR DATA ASSOCIATED WITH THE BIPHENYL MOIETY

The data set (ii): ring part, and data set (iii) are considered. They
correspond to the quadrupolar q^ and deuteron-deuteron dipolar Q.^
interactions on the two rings. These data are interpolated at
temperatures corresponding to the eight selected temperatures as defined
above.
Since all the parameters have been determined by the previous
analysis, the values of all these interactions should be predicted by
the model. Taking the values of the angles u^ determined in Section 5.1,
it is not possible to reproduce the quadrupolar data within experimental
error using Eq. (2). It turns out that agreement is obtained if all four
angles ur are allowed to increase by about ~ 0.1°. This may indicate
that the principal axis of the e.f.g. tensor acting on the deuterium
spins does not lie exactly along the CD bond. The actual explanation is
probably different. The CD direction is probably the main principal
direction of the e.f.g. tensor, but the symmetry of this tensor is not
exactly uniaxial around this bond. In this case, Eq. (2) should be
modified.
For a ring making an angle <J> with the laboratory frame Oxz (<t> = O and
- 32.21° for rings A and B, respectively), it is easy to show that we
have, for deuteron i, [16]:
1

- —

2 cos 24- (sin2u£ - y d + cos2u^)) (Tx

v sin24> < - y ( l + COS2O) LJ (3)

1
•/

83

where ^1 is the anisotropy parameter of the e.f.g. tensor acting on

deuteron i, in the frame where Oz is along the CD bond (component c. ),
Cn
Ox in the ring plane (component - y (1 - I11)), and Oy perpendicular to
c
i
the ring (component - -=- (1 + T]1)).
For the dipolar interactions, Eq.(1) is not changed.
Using these equations and fixing the angles ur to the values found in
section 5.1, it is possible to find a value of t\ for each of the four
deuterons 1, 2, 5, 6, which allows to reproduce all the data within
experimental error. More specifically, the analysis was made as follows:
the data associated with ring A were used as another manner to determine
Txx - Tyy versus temperature. Values of Ti1 and Ti2 were chosen and T2zwas
fixed according to figure 9. Consistency is achieved for values of the
I11 in the following ranges: ^=- 0.012 ± 0.005, T)2 = - 0.023 ± 0.004.
The new values of Txx- Ty y obtained in this way for the central values
of T) are reported in figure 9.
Similarly, the data associated with ring B were used as another way to
determine Txy . As above, values of T)5 and T)6 were chosen, T22 and
T xx - T were fixed (the latter being allowed to vary within the
uncertainty range). Consistency is achieved with the following values of
the Ti1. : T]5 = - 0.015 ± 0.014, Ti6 = - 0.009 ± 0.008. The new values of Txy
obtained in this way for the central values of the Ti1 are also reported
in figure 9. The quality of the fit may be appreciated in figures 5 and
6 where are reproduced -experimental and calculated values. The four
values of ^ are well inside the accepted limits ( - 0.03 < Ti1. < 0.03)
[15,16], and all of the same sign, showing the self-consistency of the
analysis.

FIGURE 5

FIGURE 6

6. RESULTS AND DISCUSSION.

f. The results of the above analysis are summarized in figures 7, 8a,b
and 9. They concern the "single" conformation through the torsional
Vj angles <f and «pj, (figure 7), the end-chain disorder (figures 7 and Sa,b),
-* and the orientational statistics of the nematic phase through the
complete tensor order (figure 9). The mirror image conformation is
obtained by changing the signs of all torsional angles.

6.1 THE CONFORMATION.

 ai
84

The conformation of t> Jiphenyl moiety, described by the angle * of

about 32°, is comparablt ;.o that found for similar molecules [6], and in
complete agreement with the value obtainr in [11]. This result will not
be commented further.
The conformation of the pentyl chain (figure 7} is more instructive.
The plane of the first CCC plane makes a large angle with the adjacent
phenyl ring. The angle <pa increases from ~ 80 to 90° with increasing
temperature, thus decreasing the internal energy until its minimum. The
next torsional angle «p^ decreases from ~ 50 to 40°, also decreasing the
internal energy, but the value is far from the absolute minimum
corresponding to O (trans). This suggests strong interaction with the
neighbours at this level of the chain. The next bond <pj. is trans and
does not change with temperature, as expected.

FIGURE 7

According to the model, this conformation should be close to that in

the solid phase. This conformation is not known since single crystal
crystal! ographi c data of 5CB do not exist so far. However some
comparison can be made with the parent molecules 3CB and 4CB. The most
interesting comparison lies in the value of <pa. This angle is about 90°
in 3CB [8], about O" in 4CB [9]. Since a value close to 90° has been
found here for 5CB, it would be interesting to check if this result is
general for the nCB series, namely if <pa is large for n odd and small
for n even. This can be only checked by single crystal crystallography.

6.2 THE END-CHAIN DISORDER.

The existence of end-chain disorder has been imposed by the model, due
to the impossibility to achieve any reasonable fit without it. This
impossibility may appear as a tiresome counter-example to the
credibility of the "single conformation" model. This is not the case,
however, because the necessity of introducing two conformations instead
•?* of one, appears to be the consequence of a structural disorder proper to
the nCB molecules, and which has been already observed in the solid
Phase of the fourth member of this family, 4CB, and in solid TBBA
- \ [13,14]. This is what will be discussed now.
As stated above, the two site model has been suggested by the results
on I:B [9]. The remarkable thing is that the two sub-models I and II
give essentially the same results, and comparison with similar results
> in the solid phase of 4CB seems to make sense:
f 0') the methyl group performs nearly the largest possible torsional
I jump. This jump, described by ^1-(P50 , is 180° (by definition) in
J sub-model I and ~ 160° in sub-model II, this latter jump angle being

JK
 'ft

•'• •> .«L •

85

practically independent of temperature. This jump is "only" ~ 60° in 4CB

[9], and is O (no disorder) in 3CB [S]. In TBBA, the overall disorder is
described by a more complex model, but the jump is roughly as large as
in 5CB.
(ii) the two valence angles U50 and usi are also similar in the two
submodels (figure 8a). It is close to the standard value (between 110
and 115° for the (P5 value farthest from the trans position ((P51= 240° or
160° for sub-models I and II), but deviates significantly from it
(between 90 and 100°) for the (P5 value closest to the trans position (
(P50= 60° or 0° for sub-models I and II) . The same result holds for 4CB.
In this compound, a "standard" value of ~ 112° is found for <p = 98.3° ,
compared to ~ 119° for (p = 41.25° (these values of (p have been
calculated using the cristallographie data of [9])- In TBBA, values of u
ranging between 95 and 105° are claimed [14].
(iii) finally the probability p (we recall that p corresponds to the
"most trans" site) is large and decreases with increasing temperature
(figure 8b). This means that the "least trans" site, tends to be
occupied more and more often as temperature increases. In 4CB, the
probability of the "trans" site ( p = 0.58 [9]) is also larger than that
of the "nontrans" site. In TBBA, the situation is described in a
different way [14] and a simple comparison cannot be made.

These results suggest a reason for such end-chain disorder. The

packing forces are such that the two sites are nearly energetically
equivalent for the internal energy. The "trans" site is favourable for
the torsional energy, but not favourable for the energy associated with
the bond angle u. The reverse situation holds for the other site. At
high temperature, the system "hesitates" between the two sites, but
since the former probably corresponds to a (slightly) lower energy, it
lives a longer time in the "trans" site. As temperature decreases, this
time increases. If the molecular arrangement is preserved in the solid
phase, it is possible that a similar disorder also exists in the solid
phase. An argument in favor of such disorder is given by the densities.
The densities of solid 3CB and 4CB are 1.354 and 1.139 g/cm3,
respectively, as calculated from the crystallographic data of [8,9]. The
large difference is consistent with the existence of disorder in 4CB.
The density of nematic 5CB at the lowest temperature in the nematic
phase is 1.01 g/cm3 [17]. The jump of the density at the solid to
1 nematic transition in most liquid crystals being less than ~ 5 %, one
can infer
that the density of solid 5CB is less than that of 4CB, in
agreement with the existence of a larger disorder. Another, but related,
argument is the melting temperature T KN , which is observed to increase
with increasing molecular mass. Indeed, for 5CB, TKN = 295.6 K is
significantly lower than for 4CB = 319.5 K, both being lower than for
• /T
3CB = 341 K which exhibits no disorder in the solid phase [8].
Conversely, the very existence of this disorder probably explains why
\ 5CB, but also 6, 7 and 8CB are room temperature nematics. As a
consequence, consistency clearly requires that analysis of similar NMR

89
 86

data in the other compounds also imply end-chain disorder.

6.3 THE ORDER TENSOR.

The last ouput of the study is the complete order tensor (figure 9).
This tensor (which, if the logic of the model is strictly followed, is
an average between two tensors: c.f. the above discussion) corresponds
to the maximun amount of information that can be extracted from the NMR
data concerning the orientational statistics of the nematic phase. It is
observed that all the components decrease (or at least do not increase)
with increasing temperature, in agreement with an overall decrease of
the orientational order. The finding that Tzz is significantly larger
than all other components show that the para-axis of the rings (Oz) is
more aligned along the director than Ox and Oy. However, since the yz
and zx components have sizable values, the "best aligned direction", or
"long molecular axis", makes some angle with the para-axis.
These results can be made more quantitative by diagonal ization of the
tensor i.e by transforming the information contained in the five T1,
components into the two principal values: the uni axial and biaxial order
parameters S2Z and Sxx- SYY, and the three Euler angles V 1 , V2 and V3,
which picture three successive rotations around the Oz, Oy and Oz, which
orient the molecular frame Oxyz in the principal frame OXYZ. The
convention adopted is th.e following: is labelled OZ the principal axis
which makes the smallest angle with the para-axis Oz, and are labelled
OX and OY the two other axes such that Sxx- SYY is positive. Figures 10
and 11 show the variation of the two order parameters and of the three
Euler angles versus reduced temperature,

FIGURE 10

FIGURE 11

All quantities are found to vary with temperature. Both order

-i
parameters decrease on heating, whereas the inclination V2 of the
'! principal axis OZ on the para-axis Oz decrease from ~ 20 to ~ 13°
*•{ throughout the nematic phase. The angle V 1 + V3 roughly represents the
'.,t angle between the OX principal axis and the Ox axis (set in the A ring).
The finding that this angle is small and negative (~ - 8°) means that
the XOZ principal plane is close to the plane of ring A, but "outside"
» the dihedral angle.
,p These results are illustrated in figure 12 which shows stereoscopic
"1 projections along the principal axes OX and OY, of the conformation
I corresponding to the middle of the nematic phase.

100
 87

1
FIGURE 12

It is observed that although the conformation is not particularly

flat, the principal plane XOZ, (by definition the plane in which the
disorder is maximum) can reasonably be considered as "the plane of the
molecule". These features are preserved over the whole nematic range,
although changing in the details.
To end up with this section, it is worth noting that the finding that
not only the order parameters, but also the principal axes change with
temperature explains the failure found in [10], of the "straight line
ratio plot representation", to describe the temperature relative
dependence of magnetic interactions, despite only one order tensor is
introduced in the model. The reason is that the interactions are linear
functions of the five temperature dependent order tensor components
(c.f.Eqs.(l) and (2)). Another way to say this is that expressing these
equations in the principal frame of the order tensor, one obtains a
linear combination of the two order parameters, whose coefficients are
temperature dependent also. These coefficients are constant only if the
orientation of the principal axes are fixed, and then, the
representation works. For more details, see [10] and references therein.

7.COMPARISON WITH RESULTS OBTAINED USING MANY CONFORMATION MODELS

Needless to say, the various NMR data considered in this work have
been already analyzed in some way, but to the best of our knowledge,
never altogether in the framework of the same model.
The data of set (i) and (iv) concerning the biphenyl moiety have been
analyzed in [10] and [11], respectively. The model used in both cases is
formally identical to the single conformation model used here, in that
sense that the mathematical expressions of the magnetic couplings are
the same, and very similar results have indeed be obtained, as expected.
Thus, even if the philosophy may not have been the same as ours, any
further discussion about this point is scientifically irrevelant at this
stage of the comparison.
More interesting are the results concerning the data associated with
the pentyl chain. Here, all the analyses made are variations of the
i "many conformation model" based on the very popular Rotameric Isomeric
, State (RIS) approximation, in which the pentyl chain can exchange among
33 = 27 different conformations. The variations Tie in different
assumptions about the coupling between conformation and orientation.
-, Historically, the first such model was presented in [18]. Each conformer
.^ is assumed to be contrained in a cylinder mimicking the "nematic
constraint", the axis of this cylinder being assumed to be either
parallel to main inertial axis (IF model) or to the para-axis (SF
model). The order tensor of each conformer is completely deduced from

Sf.
 88

* • t the inertial tensor. These two sub-models have been used to analyze the
ratios qx/ q& at one temperature, that is four data. The agreement with
experiment is reasonnable (better for IF) in view of the fact that the
only completely free parameter of the model is the radius of the
cylinder.
The second "many conformation model" is the so-called ELS model [19].
It has been used to analyze both the quadrupolar data qx and the
deuterium dipolar data Dx associated with the chain at one temperature
[19]. Here, the "nematic constraint" is pictured by the "potential of
mean torque" (HTP) which governs both the probability of occurence of
each conformation and its orientation with respect to the director. The
free parameters of the model are two internal energies associated
with the RIS model, and two coefficients in the developpment of the MTP,
that is four parameters for ten data. The result of the fit is rather
bad, as can be inferred from figures 4 and 6 of [19]. In order to
improve the agreement, the same model has been modified in [20], by
including a third coefficient in the développement of the MTP, but
dropping out one internal energy parameter, and the data concerning Dx
not considered in the fitting procedure. The situation is now five
1
parameters for five data. Perfect fits are obtained if all parameters
are assumed to be temperature dependent, but good fits are still
obtained if the single internal energy parameter is fixed. In this way,
this parameter and the three parameters of the MTP could be determined
over the whole nematic range [2O]. However, as can be inferred from
figure 9 of [20], and the error bars quoted in [10], this model does not
seem completely adequate to reproduce all the Dx within experimental
error. Finally, it may be worth noting that a strong criticism of the
model used in [18] is made in [2O]. It is indeed claimed that the
assumption made in [18] concerning the relation between order tensor and
inertia tensor is "totally unrealistic".
A. third kind of "many conformation model" has been used later to
analyze the same set of qx at one temperature [21]. Here, all the
conformers are assumed to be axially symmetric around some axis Z, and
the order is described by a single order parameter Szz, the same for all
conformers. The fractions of the conformers were determined "adjusting
the conformational statistical weight parameter of each rotatable bond
against experimental values qx". Two sub-models were used: model I in
«?* which Z is along the para-axis (same Z for all conformers) and model II
I where Z is the line joining thenitrogen of the cyano group with the
i methyl carbon (differentZ for all conformers). In each case there are
\ three parameters, namely the statistical weigths corresponding to
torsional angles ^ for A = p, -Y, 6, for four data constituted by the
ratios lqx/qj (signs are not considered). Model I appears to be very
poor, but model II yields reasonnably good results. Unfortunately the
^ predictions of this latter model for the Dx are not worked out. The
•*| authors [21] note that the distributions of the conformers they find
j resemble those found with the models of references [18] and [19,2O]. In
our
opinion, this is not surprising since all these models use the same

• --.ai

102
 89

RIS approximation to analyze the same data.

To end up with this section, it is worth noting that some analysis of
the 13C data of sets (v) and (vi) has also been performed in [12], but
there, the basic idea was not to establish models, but rather to check
if these data were consistent with the proton and deuterium data of
[10.11].

8. CONCLUDING REMARKS

In this paper, which represents the first part of the reanalysis of

data concerning (nematic) 5CB, in terms of the single conformation
model, we has shown that this model can reasonnably describe the full
set of presently available NHR data. It must be realized that if all the
experiments have had been performed at the same temperature, the
computational situation would have been twelve free parameters to be
determined with fourty data. Not all these data can however be
considered as independent.
' The most important result of this study is probably the apparent
i failure of the model in its strict sense to describe all the data. We
have seen that far from being negative, this result may be considered as
a support for the basic idea underlined in the model, in that sense that
(i) it predicts the existence of a subtle phenomenon within the
molecule, namely end-chain disorder and (ii) it suggests that
application of the model in its strict sense may apply to an entity
larger than the single molecule, possibly to two neighboring (and
enanthiomeric) molecules, or even to the 2 N molecules of the unit cell
of the crystalline phase. This possibility was suggested to us by the
the detailed crystallographic study of TBBA [14].
Concerning point (i), support for end-chain disorder in 5CB is so far
indirect, since it relies on comparison with what occurs in similar
molecules and considerations about molar volumes and melting
temperatures. The strongest support would be provided by a direct
evidence of such disorder in the solid phase, either by single crystal
crystallography, as for 4CB and TBBA, or possibly by (very) high
quasi-elastic neutron scattering. Concerning point (ii), the situation
is presently only speculative since it cannot be tested experimentally
S** on this system, and by NMR, since all intermolecular magnetic
'j interactions are averaged out to zero by the translational motions.
1
J A few words about the "many conformation" models are in order.
- \ Clearly, they cannot be ruled out since a similar complete study has not
be performed with them. However, it seems possible to exclude the kinds
of models, such as those of [18-20] from the class of physically
plausible models (and put them in the class of operational models
V applicable to the analysis of NMR data of liquid crystals), simply on
the basis of thermodynamical results concerning the clearing enthalpies
and entropies: ASNI= £HNI/ TNI. The argument has been applied in [2] in
relation with NMR data of a dimer liquid crystal, but also applies to

•3!..

103
 .4*. «r.
90

5C6 as shown below.

We recall that the basic ingredient of these models is that the nature
of the phase not only influences the long range order, but also the
probability of occurence of the conformations. More specifically, the
conformations are constrained by the "cylinder" in [18], or by the
"potential of mean torque" in [19,20], whereas, in the isotropic phase,
the molecules are free to adopt all the possible conformations as in the
gas phase. Thus, these models predict that the values of the above
quantities come from two additive contributions: a conformâtional and an
orientational contributions. In [22] are calculated the conformational
contribution to the energy (K enthalpy if the small volume change is
neglected) and to the entropy changes for the models of [20,21] and
[22]. These values are 0.202 kcal/mol and 0.939 R for [20,21], and 0.143
kcal/mol and 0.800 R for [22]. The experimental values are 0.12 kcal/mol
and 0.200 R [22], that is much lower, at least for the entropy, than the
conformational contribution alone! The authors of [22] suggest an
explanation for the discrepancy in the entropy which is completely
incomprehensible to us (and to several scientists that have been
consulted), and "have no plausible explanation" for the discrepancy in
the enthalpy.
The qualitative explanation is very simple in the framework of the
solid like-picture of a liquid [4]: there is no essential differences
between local properties in all condensed phases. So, all conformational
(i.e molecular) contributions to thermodynamical changes at transitions
between condensed phases should be small, presumably negligeably small
between two liquid phases such as the nematic and isotropic phases.
The analysis can be made more quantitative using the single
conformation model and the results of the present study. According to
this model, the molecules can be considered as rigid, so theories as
simple as the Maier and Saupe theory should be sufficient to predict the
order parameter dependence of the enthalpies and entropies. The
prediction is that both quantities scale as S^1, where SNI is the order
parameter at the clearing transition, to be identified with S 2ZiN1 if the
order is biaxial, since this quantity pictures the maximum total
orientational order (or disorder). From figure 10, we estimate Szz NI «
0.36.
All these numbers can now be compared with those of, for example,
if» para-azoxyanisole (PAA), which can be considered much more rigid than
1
I 5CB, since it contains no true flexible chains. For PAA, AHNI ranges
between 0.14 and 0.18 kcal/mol and ASNI between 0.17 and 0.22 R [22]. On
the other hand, Szz NI is about 0.35 [2,23]. This value is the same as
that of 5CB, if one considers that the uncertainty on absolute values of
order parameters is not less than 5%. If so, the two thermodynamical
quantities should also be the same for both compounds. It is seen that
> this prediction is excellent for the entropy, and acceptable for the
enthalpy. In any case, one cannot reasonnably expect perfect agreement
in view of the roughness of the theory used.
To end this discussion, it must be said that if these results appear
 91

to constitute a refutation of the class of "many conformation models"

which imply that the probability of occurence of the conformations is
fundamentally affected by the nature of the phase, such as the models of
[19-21], they say little about the question: do , in a fluid condensed
phase (isotropic, nematic or other), flexible molecules exist in
essentially one conformation (solid-like picture of a liquid) or in
several (very) different conformations (gas-like picture of a liquid).
In this respect, provided that it will prove to describe satisfactorily
the same amount of data, the model of [22] is acceptable if the idea
that, in the isotropic phase, the molecules behave as if they were
isolated is abandonned. In fact, this latter assumption is not
necessary: it is sufficient to assume that the order parameter S22
introduced in this model is zero in the isotropic phase.
So far, the question of which picture is better is still open, only
arguments in favor of the former have been provided by us in [1-6], and
clearly, in this work also (despite the failure of the single
conformation model in its strict sense). Further arguments are given in
the next paper of this issue, in which it is shown that the results of
the present study allow to describe satisfactorily a number of data
other than NMR.

10. ACKNOWLEDGMENTS

The authors are highly indebted to Professor B.M.Fung for detailed

correspondence concerning his work published in [12], and his kind
collaboration. This work was partially supported by NSF/CNRS Grant INT -
871501/88 - 920156.

Vl

1
• f'-f

92

REFERENCES AND FOOTNOTES

[1] Galland D. and Volino F-, J.Phys.France 50 (1990) 1743.

[2] Volino F., Ratto J.A., Salland D. and Esnault P., MoI .Cryst.Liq.
Cryst. 191 (1990) i23.
[3] Galland D. and Volino F., J.Phys.II France I (1991) 209.
[4] Volino F. and Galland D., Proceedings of the European Conference on
Liquid Crystals, Courmayeur, Italy (1991); MoI.Cryst. Liq. Cryst.
212 (1992) 77.
[5] Galland D., Gérard H., Volino F. and Ferreira J.B.,
Macromolecules (1992), in print.
[6] Gérard H., Avalos J., Galland D. and Volino F., Liquid Crystals
(1992) in print.
[7] a) Cailleau H., Baudour J.L. and Zeyen C.M.E., Acta Crys. B35.
(1979) 2704.
b) Cailleau H., Baudour J.L., Meinnel J., DworkinA., Moussa F.
and Zeyen C.M.E-, Faraday Discuss. Chem.Soc., 59 (1979) 7.
[8] Haase W., Paulus K. and Pendzialek R., MoI.Cryst. Liq.Cryst.,
100 (1983) 211.
[9] Vani G.V., MoI.Cryst.Liq.Cryst., 99 (1983) 21.
[10]Emsley J.W., Luckhurst G.R. and Stockley C.P., MoI.Phys. 44 (1981)
565.
[ll]Sinton S.W., Zax D.B., Murdoch J.B. and Pines A., MoI.Phys. 53
(1984) 333.
[12]Fung B.M., Alfzal Jalees, Foss T.L. and Chau Mei-Hing, J.Chem.Phys.
85 (1986) 4808.
[13]Volino F., Dianoux A.J., Lechner R.E and Hervet H., J.Phys.France,
Colloques 36 (1975) Cl-83.
[14]Doucet J., Mornon O.P., Chevalier R. and Lifschitz A. Acta Cryst.
B33 (1977) 1701.
[15]Emsley J.W. and Lindon J.C. "NMR Spectroscopy Using Liquid Crystal
Solvents", Pergamon Press (1975).
[16]Doane J.W., in "Magnetic Resonance of Phase Transitions", Owens
F.J., Poole C.J. and Farach H.A., Academic Press (1979).
[17]Dunmur D.A and Miller W.H., J.Phys.France, Colloques 40 (1979)
C3-141.
[18]Samulski E.T and Dong R.Y., J.Chem.Phys., 77 (1982) 5090.
[19]Emsley J.W., Luckhurst G.R. and Stockley C.P., Proc.R.Soc. A 381
(1982) 117.
[20]Counsell C.J.R., Emsley J.W., Heaton N.J. and Luckhurst G.R.,
MoI. Phys., 54 (1985) 847.
[21]Abe A. and Furuya H., MoI.Cryst.Liq.Cryst., 159 (1S88) 99.
[22]"Source of Thermodynamical Data on Mesogens", MoI .Cryst.Liq.Cryst.,
115 (1984) n° 1 to 4.
[23]Galland D. and Volino F., J.Phys.France, 50 (1989) 1743.
 : i.
t '</

93

TABLES

Table 1;_ Proton-proton dipolar couplings D1^ associated with the

biphenyl moiety of 5CB at T - TNI= - 11 K. The observed values are taken
from [U]. The calculated values are those for the single conformation
model. See text for details.

i,j Observed Calculated

1,2 - 4460 (- 4478) - 4460

<1,4> 377 ( 385) 398
1,3 72 ( 47) 48
5,6 - 4500 (- 4478) - 4500
<5,8> 388 ( 385) 388
5,7 70 ( 47) 46
1,5 - 150 ( - 147) - 143
<1,6> - 385 { - 365) - 398
2,6 - 1721 (- 1741) - 1721
 -' -y'

ai

94

Table 2._ 13 C -dipolar couplings D1d of 5CB at T - T NI = - 5 K. The

observed values are taken from [12]. The calculated values are those for
the single conformation model.

i,j Observed Calculated

IM 0.81 ± 0.04 0.804

IV - 1.15 ± 0.04 - 1.134
2', 2 1.08 ± O. 04 1.071
2M - 1.15 ± 0.04 - 1.130
5', 5 1.03 ± 0.04 1.028
5', 6 - 1.10 ± 0.04 - 1.132
6', 6 1.02 ± 0.04 1.023
6', 5 - 1.12 ± 0.04 - 1.132

a',a 4.13±0.16 4.147

p',p 2.75 ± 0.12 2.781
T',-Y 2.96 ± 0.12 2.992
&',6 1.98 + 0.08 1.988
u',to 1.39 ± 0.06 1.450

IÎ*

'1

1
1
J/

95

UJ

Fig.l Sketch of pentyl-cyanobiphenyl (5CB) molecule ghowiag the

labelling of the various spins and the bond angles and dist^t«=es used in
the study. The molecular frame set in ring A is also shown.
»1
 96

B1, (Hz)

-20 -16 -12

1
I
S?*

Fig.2 Proton-proton dipolar couplings D 1 2 , D13 and D14 of 5CBdIS versus

reduced temperature. The triangles are the measured values of [1O]. T-e
continuous curve joins calculated values. Note that the fit has been
made using experimental values interpolated to temperatures
corresponding to the deuterium data of the same referci.ce. Error bars
claimed in [10] are shown.

.
1
 97

Fig.3 Quadrupolar couplings associated with the deuterons of the pentyl

chain of 5CBdl5 versus reduced temperature. The points are the measured
values of [1O]. The continuous line joins the calculated values.
Experimental error bars are within point dimensions.
 98

-20 -16 -12 -8

Fig.4 Deuterium-deuterium dipol^-v couplings associated with méthylène

Vl..* and methyl groups of the pentyl chain of 5CDdIS versus reduced
temperature. The points are the measured values of [10]. The lines join
the calculated values. Error bars claimed in [10] are shown.
 q,- (kHz)
12

10

-16 -14 -12 -10 -8 -6 -4 -2 O

[
Fig.5 Quadrupolar couplings associated with the phenyl deuterons of
5CBdIS (qR) and 5CBdS Cq1 and qs) versus reduced temperature. The points
are the measured values 110] and the lines join the calculated values.
Error bars are within point dimensions. Note that qR corresponds to the
average between qs and q6 whereas qs is the average between q2 , q5 and
qs . Individual values cannot be discriminated experimentally with the
,if two differently deuterated substances.

}
 100

(Hz)

140

120 56

100

80

60

40

20

-16 -14 -12 -10 -8 -6 -4 -2 O

f
i Fig.6
rings
Dipolar couplings between deuterons in ortho positions on phenyl
A and B versus reduced temperature. D12 is from the data with
5CBd8 and D56 is from the data with 5CBdIS. The points are experimental
(note the large error bars) [10], and the lines join calculated values.

\
 SfL .

101

VxO
ziu-
180
% ^ * + * * * * *
150 ' 1

120
90 • c p . • . • •
tt « • •
60 Cp* * * » * •» * *
5
Q
30
O . q >Y
-30
p • •
-60
.on • • • • i • • • •

-20 -16 -12

Fig.7 Torsional angles «p^ of the pentyl chain of 5GB versus reduced
V! temperature predicted by the single conformation model. The two Cp5
values correspond to the two-site sub-models I and II used to describe
the end-chain disorder. Note the weak and continuous changes over the
entire nematic phase.

r./-
 102

M')
120

110

U6
!! o0o
100

90
-20 -16 -12 -8 -4 O

.75 • •° *
° o
* « o
S

.25

-20 -16 -12 -8 -4 O

J
1
Fig. 8 Parameters of the two-sitesub-models I and II used to describe the
end-chain disorder, predicted by the single conformation model, versus
reduced temperature.
a) Values of the two bond angles us (uso and U51 ;. The two values for
4CB [9] , and the range of values for TBBA [14] are indicated.
b) Probability p of the "cost trans" site (index O). Note that this
\ probability decreases with increasing temperature.
 103

.6

.4

.3

.2

A"yy A Aa A AA
8 :
1Vl
5!*.
-20 -16 -12
Fig.9 Components of the order tensor in the Oxyz frame, predicted by the
single conformation model, versus reduced temperature. Two sets of
values for Txx- Tyy and Txy determined using the proton data (larger
symbols) and the deuterium data (smaller symbols) are shown. See text
for details. The dispersion on values of Txy pictures the maximum error
bar on all components.
 104

.7

.6

-St

.4

.3

.2

.1

O
-20 -16 -12 -8

Fig.10 Uniaxial and biaxial order parameters of nematic 5CB versus

reduced temperature, obtained by diagonalization of the order tensor.
Hote that both parameters decrease with increasing temperature.
 105

60 V,
3

40

20

-20

-40

-60

-80
-20 -16 -12

Fig. 11 Euler angles V1 , V2 and V3 which describe the orientation of the

molecular frame Oxyz in the principal frame OXÏZ. They correspond to
successive rotations around Oz, Oy and Oz. Note that the inclination V-
of the para-axis Oz of the rings on the principal axis OZ decreases on
heating.
 -'fe

106

X Y

JX Y

Fig.12 Stereoscopic projections along the principal axis OY (upper) and

along the principal axis OX (louer) of the order tensor, of the
conformation of 5CB predicted by the single conformation model. This
conformation corresponds to the middle of the nematic phase. The
end-chain disorder is pictured by sub-model 1 (c.f. section 5.2). The
cyano group is not represented for simplicity.

;*>
\

120
 107

E B 2 Exploitation des résultats provenant de la RMN

Les résultats de l'analyse des données de RMN du 5CB, obtenus

sous l'hypothèse d'un désordre en bout de chaîne et dans le cadre du
modèle à conformation unique, ont été confrontés à un certain nombre
d'autres données, associées à des propriétés tensorielles de second
ordre, et pouvant donc s'exprimer à partir du tenseur d'ordre (de rang
2) déterminé. Le but de cette analyse est de vérifier la validité du
modèle appliqué à des données autres que RMN et d'estimer des
quantités moléculaires physiques associées aux données considérées.
Toutes ces données correspondent à des mesures selon le directeur
(ces données seront indicées II) ou perpendiculairement à celui-ci
(indice ±). Les valeurs observées souscrivent à la formule (13), qui
les relient au tenseur moléculaire Q de rang 2 associé à la propriété
observée et au tenseur d'ordre T qui décrit l'ordre statistique
moléculaire vis-à-vis de la direction d'observation. Dans le cas des
interactions magnétiques, on a déterminé T = T(( qui décrit le désordre
par rapport au vecteur directeur n0. Pour une direction de mesure n
différente de n0 , on doit alors considérer le tenseur T(n) . Si la
statistique de l'axe principal Z de la molécule est uniaxiale par
rapport à n0, on peut alors montrer que pour n ± n0 on a Tx = - •=• TU .
En combinant les deux équations obtenues selon la formule (13),
l'anisotropie Q3 et la valeur moyenne Qav du tenseur Q sont alors
données par:

= «II - Qj. = Tr(QT) (26)

6t
„ 1. 2. 1
J QII + | Q1 = j Tr(Q) (27)
Les données exploitées ont permis de déterminer le tenseur x de
susceptibilité magnétique, le tenseur de polarisabilité a ainsi que le
moment dipolaire permanent y. de la molécule. Les composantes des
tenseurs « et x sont des quantités moléculaires moyennées sur tous les
mouvements internes à la molécule, dont les mouvements liés au
^ désordre de bout de chaîne. Dans l'absolu, les données de RMN ayant
*• « montré que la conformation de la chaîne pentyl variait, quoique
! légèrement, en fonction de la température, ceci devrait influer sur la
v} valeur des deux tenseurs et de ^ , et l'on devrait considérer un
.- tenseur (légèrement) différent à chaque température pour chacune des
propriétés.valeurs effectives des tenseurs x et a. Cependant,
préférant omettre ces (faibles) variations conformationnelles afin de
»., limiter le nombre de paramètres du problème, deux uniques tenseurs x
\ et a ainsi qu'un unique moment dipolaire ont été considérés pour
'T l'ensemble des températures. De même que dans le cas du tenseur
1 d'ordre, on n'a considéré qu'un seul tenseur pour les deux

121
 108

conformations terminales possibles.

Cette analyse de différentes données concernant le 5CB en phase

nématique selon les principes du modèle à conformation unique s'est
révélée extrêmement fructueuse (dans le cas des données RMN, on a
reproduit à huit températures dix-neuf données expérimentales grâce à
douze paramétres), le principal résultat étant son caractère prédictif
quant à un désordre en bout de chaîne aliphatique. Avec cette
hypothèse, les données ont pu être analysées donnant des informations
sur la géométrie de la molécule, l'ordre au sein de la phase nématique
(le tenseur T), la connaissance de cet ordre permettant par la suite
d'obtenir de nouvelles informations sur les tenseurs moléculaires de
susceptibilité magnétique et de polarisabilité.
Bien sûr, toute ces informations dépendent de l'existence de ce
désordre terminal. S'il s'avérait qu'un désordre n'existe pas dans la
phase nématique, le modèle à conformation unique devrait alors être
abandonné pour des molécules trop complexes. Si, par contre, ce
désordre était mis en évidence, par des moyens autres que l'analyse de
données RMN, dans cette phase nématique, voire dans la phase solide du
5CB, le modèle à conformation unique s'imposerait en tant que moyen
d'analyse puissant des données RMN en phase nématique puisqu'outre ses
informations précises sur la structure et l'orientation des molécules
au sein de cette phase il est capable de prédire l'existence de
phénomènes plus subtils telle cette délocalisation des carbones
terminaux de la chaîne alîphatique.

122
 109

THE MOLECULAR PHYSICS OF NEMATIC PENTYL-CYANOBIPHENYL (5CB)

REVISITED IN TERMS OF THE SINGLE CONFORMATION MODEL:
II-APPLICATION TO THE ANALYSIS OF DATA OTHER THAN NMR

H.Gérard, D.Gall and and F.VoIino*

DRFMC/SESAM/PCM - CENG - 85X - 38041 F-Grenoble Cedex

ABSTRACT

The results of reference [1] concerning the conformation and the order
tensor of nematic pentyl-cyanobiphenyl (5CB), deduced from the analysis
of a large set of NMR data in terms of the single conformation model,
are used to analyze the magnetic susceptibility data of [2-5], the
optical refractive index of [5,6], the electrical permittivity data of
[7], and the magnetically induced birefringence data of [8]. The
following molecular quantities are deduced: the complete (i.e. principal
values and principal directions in a molecular frame) magnetic
susceptibility and optical polarizability tensors, and the complete
(i.e. modulus and orientation in the same frame) permanent electrical
dipole moment. These results do not reveal any contradiction with well
established results. The. molecular Cotton-Mouton constant is also
deduced, and its use in the analysis of the data of [8] suggests that
the short range orientational correlations between neighboring molecules
extends far inside the isotropic phase. Comparison with previous
analyses of the same data is made. The efficiency of the single
conformation model is emphasized, in that sense that not only it can
describe accurately data for a number of molecules, but that its failure
in its strict sense in case of 5CB is the signature of a particular
phenomenon at the molecular scale. Some considerations about the
generalisation of the model are also made.

v
Member of CNRS

'I»•{ Short title: MOLECULAR PHYSICS OF 5CB REVISITED: NON NMR DATA

Classification Physics Abstracts: 61.16N 61.30

123
 • •-. ,&L

110

1. INTRODUCTION
In the preceeding paper of this issue [1], we have presented the
results of the analysis of a large set of NMR data concerning the
nematic phase of pentyl-cyanobiphenyl (5CB), in terms of the single
conformation model. This model could explain all the data considered,
provided the existence of end-chain disorder, corresponding to
délocalisation of the methyl carbon, was assumed. Several indirect
arguments in favor of this disorder were given. The conformation and the
complete tensor order were determined, and a simple model to describe
this disorder was proposed. It was found that the (single) conformation,
and the orientation of the principal axes of the order tensor slightly
change with temperature. Both the uniaxial and biaxial molecular order
parameters were found to decrease with increasing temperature.
In this paper, we use all these results to analyze a number of other
data, focussing on data associated with second rank tensorial
properties, which relate directly to the (second rank) order tensor.
Since this tensor has been completely determined, these analyses are
used to check the single conformation model and/ or estimate molecular
physical quantities relevant to the particular data considered. A number
of new results are obtained, and all are consistent with what could have
been expected a priori. The approximation that is be made is that the
slight changes of the conformation over the nematic phase, have
négligeable effect on the molecular tensors, so that they can be assumed
to be temperature independent. This proves to be highly sufficient to
describe all the data considered.

2. THE DATA

The following published data sets will be analyzed

Set (i) the magnetic susceptibility data of four studies [2-5]

Set (ii) the optical birefringence data of [5,6]
Set (iii) the electrical permittivity data of [7]
Set (iv) the magnetically induced optical birefringence data of [8]
s*
8*

•I 3. BASIC EQUATIONS
In
\ [I]. theoretical expressions for average magnetic interactions have
been given. All of them are particular cases of the following formula:

» Qn n = 4 Tr(Q) + jTr(Q.T) (1)

•> - J d
J wher
H e Q nn is the statistical average of the projection, along some
\ direction n of the laboratory frame, of any molecular second rank tensor

124
 -- -- . - .-.„.!"» I1 _

111

Q. The order tensor T (noted S when defined in its principal frame)

describes the orientational statistics of the molecules with respect to
n. For a uni axial nematic, T is generally defined with respect to the
director n0 . Finally the symbol Tr means the trace (spur) of the
corresponding tensor. All equations used in [1], which describe
projections of magnetic interaction tensors along the static magnetic
field H0 (basically, the direction along which the information is read)
correspond to the situation where H0 is along n0.
If the reading direction n is different from n0 , one must replace
T (sT("o)) by T<"> in Eq.(1). If the statistics has uniaxial symmetry
around n0, and 8 the angle between n and n0, it can be shown that T(n)=
P2(cos9) T .The data sets considered in this paper correspond to
measurements along the director (as in [I]), that will be labelled by
the index II, and perpendicular to the director, index j.. The two
relevant order tensors are thus TU = T and T^ = - ^-T. Combining the
above formulae, the anisotropy Qa and the average Q av of tensor Q are
given by:

Q3= Qn - Qx = Tr(QT) (2)

Qav -j Q,| + ^ Q 1 - y Tr(Q) (3)

Of interest is also the following quantity Z^, which is less dependent

on (often ill-known) absolute values of numerical constants:

Tr(QT)
v = ^= 3
Tr(Q)

We now use these expressions to analyze the above mentie-sd data sets.

4. MAGNETIC SUSCEPTIBILITY DATA.

In this section, we show that knowledge of the complete order tensor

[1] can be used to estimate all the components of the molecular magnetic
susceptibility tensor x from susceptibility measurements.
Four sets of equivalent data are published in [2-5]. These data are
T, given in the form of the values of the anisotropy of the magnetic
I susceptibility xa = XK - Xx versus reduced temperature. In each case,
si they can be represented very accurately by an empirical function (a
I Power law of the form (T - T 0 ) r , where y is an exponent smaller than 1/2
and T- a temperature slightly higher than T N I ) , given the authors. For
our purpose, these functions are used to "calculate" the experimental
t values of Xa at nine selected (according to [I]) temperatures in the
V nematic range.
\ In an ideal situation, the four laws would be identical. In fact they
are not
\ (except those of [3] and [5]), and comparison between them

125
 112

constitutes an estimate of the systematic errors inherent to this kind

of measurements. The three different data sets will be considered
successively on the same footing. All the data need to be expressed in
the same units. For convenience, we choose (c.g.s.) units of 10"6 cm3
per mole of 5CB (1 cm3 = 10~7 O.G'2). Expanding the trace in Eq.(2) in
the Oxyz frame defined in [1], it is easily shown that we have:
... = TV - /v 4- y Ï 1 T 4- — i Vx x ~ V J l T ~ T J +
Aa LA22 9 **Scx ^y y ' J zz 9 yy * xx y y'

2 xxyTxy + 2 xy2Tyz + 2 X2xT2x (5)

where x, (i>J = *,y,z) are the components of x if the Oxyz frame. This
equation is completely general. For our model, the X1 j are in fact
quantities averaged over the internal motions. Since the order tensor
components are known [1], the data may be used to determine the five
linear combinations of the X1J (assumed to be independent of
temperature), contained in Eq.(5) and/ or provide a further test of the
model. The values of xa at the nine temperatures mentioned represent
nine equations, which can be solved using a mean-square fitting
procedure. It turns out that, because Txx- Tyy and Txy are relatively
small (see figure 9 of [I]), the uncertainty on the best fitted values
of the corresponding coefficients is very large.
To reduce this uncertainty, we make the (possibly questionable)
assumption (suggested in.[5]), that the principal planes XOZ and YOZ of
tensor x are close to the two bissector planes of the cyano-biphenyl
moiety. This is rigorous for the biphenyl molecule, for symmetry
reasons. This condition is mathematically equivalent to the constraint
that Xxy is close to j (Xxx- xyy) tan*, where 4> = - 32.21° [1], is the
dihedral angle between the two rings . This constraint is not sufficient
however, since satisfactory fits are obtained for any value of xxx- X
between ~ O and ~ 160. We thus used the data of [9], cited in [5],
concerning single crystal data of biphenyl molecule. For this molecule,
the principal axis OZ of x is along the para-axis and OX, OY in the two
bissector planes (OX in the small angle). The principal values are xzz =
- 63.4, XXX = - 63-25« XYY= ' 182.15, yielding Xxx- XYY = 118-9> a11
expressed in the above units, per mole of biphenyl. If the following
assumptions are made for 5CB (i) the contributions of the cyano group
and of the pentyl chain have nearly cylindrical symmetry around the
para-axis of the biphenyl moiety and (ii) the principal axis OZ of its
susceptibility tensor makes a small angle with the para-axis Oz, then we
can estimate that the component xxx- xyy(in our reference frame) is
COS2*
close to 118.9 = 86.6 (per mole of 5CB). The correction factor
takes into account the fact that the dihedral angle between the rings is
~ 10° in solid biphenyl [1O]. With this further constraint, good fits
are obtained, for all sets of susceptibility data of [2-5]. The
corresponding values of the six components, calculated with

126
 _ _ i _ L •-~.'- - -,'.4-W** (W* _

113

xav = T (xxx+ xyy+ x*z) = - 167.3 [5], are given in Table 1. The quality
of the fits obtained can be appreciated on figure 1.

TABLE 1

FIGURE 1

Diagonalization of tensor x yields the following principal values and

Euler angles (same conventions as in [I]) . The error bar mentioned
corresponds to averaging over the four sets of results of Table 1.

XZZ = - 130 ± 3 XXX = -139 ±3 XYY = -233 ±3

V1 = - 6.6 ± 2 ° V2 = - 17.6 ± 3 ' V3 = 21.8 ± 3"

Figure 2 shows two stereoscopic projections of the molecule along the

two principal axes OX and OY. It is observed that the para-axis lies
practically in the YOZ principal plane, at an angle IV 2 I which is not
particularly small, and that the pentyl chain is roughly in the YOZ
plane and makes a large angle with OZ. Further, the principal values are
significantly different from those of biphenyl, given above. This means
that the contribution of the pentyl chain to the susceptibility is not
small, and consequently the two assumptions which have allowed to
determine completely the tensor, may not be very good. Being not
specialists of these matters, we cannot comment further on the present
results. The conclusion is that we cannot claim that the results
concerning the x tensor are unequivocally established, since they depend
significantly on the two above assumptions. The only result which is
nearly independent of these assumptions is the value of the principal
value Xzz (and clearly also of Xzz- xav)-

-i FIGURE 2

5. OPTICAL BIREFRINGENCE DATA

\f_ Optical birefringence data of 5CB are reported in [5] and [6]. The
^j analysis of these data can be made in a very similar manner as above,
the magnetic susceptibility tensor x being replaced by the
\ polarizability tensor a. Defining the optical birefringence In as the

127
 ..-.. . .. • -w ,»:
114

ratio (nf, - n^)/(nav- 1), where the n are refractive indexes parallel
and perpendicular to the director, and averaged, it is easily shown we
have (see also [5]):
^ _ C !_ . U T +i( (T T )+
2ttxyTxy + 2 a y z T y 2 + 2 ^ x T 2 x ] (5)

where a.^ are the components of a (at the wavelength of the

experiment), aav is the one third the trace of this tensor, and C is a
constant close to unity which takes into account internal field effects.
The two values C = I . (Vuks correction) and 1.24 (Neugebauer correction)
have been proposed [6], and both will be considered in the analysis.
Values of Zn measured at A = 6328 A are given in Table 1 of [S] versus
temperature. These data can be represented with high accuracy by the
same kind of empirical law as for the magnetic susceptibility data, and,
as above, this law was used to "calculate" the experimental values of Zn
at the same nine temperatures.
The same fitting procedure can be used to determine the five
parameters of Eq.(5), and the same kind of undeterminacy occurs if some
information concerning the tensor a is not injected in the analysis.
Such information exists and concerns aav [6] and the mean square
polarizability -v2 [11]. At this stage, the a^ need to be expressed in
some common unit. We choose the popular (c.g.s.) units of 10~24 cm3 (=
A3 ) per molecule. Note that a more rational unit, comparable to that
used in magnetic measurements, would be cm3 per mole (1 cm3 = 1.1264
x IQ-16 C2In2J-1).
The values of aav at the wavelength of 6328 A is 33.2 A3[6] (note that
the value at audio frequencies is very close: 33.7 ± 0.3 [7]). The mean
square polarizability <i2>, given by [U]:
2 , 2 r, 1, ,,_ 1r ,, , . ,
— cv~> = — [(a 2z - •=• (oe xx + a )]^ + — [a xx - a ] + 2 [a2 + a2 + a „] (6)

has been measured [11], at the same wavelength, for several nCB
molecules, namely for n = 2, 6, 8, 9, using depolarized Rayleigh light
scattering from solutions in carbon tetrachloride. Examination of these
data suggests that for 5CB, the value of y = <T*> 1 / z lies between 21.4
and 21.5. In the fitting procedure, we have imposed that the value of -y
given by the above equation lies between these limits.
With these two additional conditions, unequivocal results are
obtained. The values of the six components of ce deduced from this
analysis, for the two possible values of the constant C, are given in
Table 2. The quality of the fit can be appreciated in figure 3.

TABLE 2

128
 . - , - . . . -.--:- ' " —«.ui iWl; r

$?**^ ' '

115

FIGURE 3

Figure 3 shows that good fits are obtained for the two cases, and the
values of the tensor components very close to one another, showing the
small effect of the internal field correction on the results.
Diagonalization of tensor a yields the following principal values and
Euler angles, for the two sets of results of Table 2 (in parenthesis for
C = 1.24)

«zz = 46.41 (44.03) Oxx= 31.47 (35.83) c^= 21.72 (19.74)

V 1 = - 66.06 (-136.03) V2 = 7.36 (7.04) V3 = 41.05 (109.17)

It is observed that the inclination V2 of the principal axis OZ on the

para-axis is small (7.2 ± 0.2°) and significantly smaller than for the T
and x tensors ( ~ 15 and 18°, respectively). The angle V 1 + V3= - 26 ± 1°
being negative, this means that, contrarily to the situation for x» the
XOZ principal plane of a lies outside the inner dihedral angle of the
rings. This result can be appreciated in figure 4 which shows two
stereoscopic projections of the molecule along the principal axes OX and
OY.

FIGURE 4

These results concerning tensor a appear to be in agreement with what

is expected. The principal axis OZ is close to the para-axis, and the
uniaxial anisotropy Axu = «zz- y Kx+ 0^y) = 18 ± 3, is in perfect
agreement with independent determination in the audio frequency range (
A**,0'= 17.6 ± 0.8 [7]). The result concerning the biaxial anisotropy &xb
= ŒXX- O=YY = ^- 3 is more difficult to discuss since there is no
theoretical or experimental result with which it can be compared. The
only thing that can be said, is that the orientations found for the
principal axes OX and OY are consistent with the fact that phenyl rings
are more polarizable in their plane than in the perpendicular direction.
This can be inferred from detailed observation of figure 4.

\
 116

7. ELECTRIC PERMITTIVITY DATA.

Electric permittivity data of 5CB, in the audio frequency range (1592

Hz), are reported in figure 1 of [7]. They are given in the form of the
relative permittivity e» and €x versus reduced temperature, but values
are given only over ~ 10°C inside the nematic range. The physics is
similar as in the optical range but with three complications:
(i) the permanent dipoles now contribute to the induced electrical
moment.
(11) the static (or audio) polarizability tensor at0) is a priori
different from the tensor at optical frequencies a.
(iii) two kinds of corrections need to be made: the so-called cavity
field and reaction field corrections, which are strictly speaking
anisotropic.
The importance of these corrections can be appreciated on the
temperature dependence of the average permittivity e av . It is observed
[7] that £av increases by about 10% with increasing temperature. Such an
increase is clearly due to these correction factors only, since the two
natural causes for the variation, namely the density and the value of
temperature itself (via the Boltzmann factor) tend to decrease e av .
Incidently, the observation that the average refractive index decreases
with increasing temperature [6,7], shows that these correction problems
are presumably minor in the optical range.
We have attempted to make analyses using anisotropic correction
factors as in [7], and it turns out that although different in details,
the results obtained are similar to those assuming isotropy. Since it
seems that, anyway, there is not complete consensus between specialists
of how these corrections should be made, and that our purpose here is
not to discuss the theory of the electric permittivity of nematics, we
shall present here the results assuming isotropic correction factors,
keeping in mind the limitations mentioned. It turns out however, that
the results are rather stable with respect to these corrections.
According to Eqs(l-4), and the standard equations for the permittivity
[7], we have:

, Tr[(tt<°)+ F A)T]
f, ^ E^T-3 Tr(aCO)+FA) (?)

s.i where a(0) is the static electrical polarizability tensor, A is the

I following tensor constructed with the permanent dipole moment n,:

;\. = n, Uj/ kBT (8)

y F the (isotropic) reaction field correction factor, whose effect is, in

'? practice, to increase (since F is typically of the order of 1.4) the
effective magnitude of ^.
The data (first member of Eq.(7)) are extracted from figure (1) of

130
 117

[J]. For the second member, (i) the static tensor a(0)has been replaced
by the tensor at optical frequency^a, „ w h i c h h a s been completely
determined in the previous section (the atomic^contribution appears to
be small compared to the!electronic one: see above) (ii);the magnitude
of n has been fixed to two possible values reported in the literature,
obtained from permittivity measurements in dilute solutions in apolar
solvents. These are p. =4.76 ± 0.05 Debye in benzene [7], and 4.11 ± 0.45
Debye in carbon tetrachloride [12] (1 Debye = 3.33564 x 10'30C m). These
last results show that the uncertainty on \L is of the order of the
correction factor. Thus, for simplicity, we have performed the analysis
with the two possible values of p., and with tensor a of Table 2,
assuming F = I . The two free parameters of the fit are the polar and
azinmthal angles 6^ and ip^ of ii in our molecular frame Oxyz, to be
determined with five data, corresponding to five selected reduced
temperatures in the range O, -10 K. Very similar fits are obtained with
the four combinations of values of a and M,. The values of the angles are
very stable : 9^ = 31.9 ± 0.3°, and ^= 66.4 ± 2.8". The quality of the
fit can be appreciated in figure 5. It is observed that the fits are
reasonnable, except very close to the clearing point where the
calculated value is systematically too large. It is not possible to
decide whether this comes from a limitation of the model or from
experimental uncertainties which cannot be very small in this
temperature range (changing the value of F does not provide significant
improvement).
More interesting is probably the result itself, as pictured in figure
6 which shows two complementary projections of the molecule
perpendicularly to the dipole moment. It is seen that \i makes a
significant angle (O11) with the para-axis, but more interesting, it lies
roughly in the plane containing the para-axis and the average pentyl
chain, that is, close to the "elongation axis" of the molecule. This
orientation is broadly consistent with what is known concerning the
orientations of the dipole moments of cyanobenzene and pentylbenzene
molecules, assuming additivity. As for tensors * and a, it thus seems
that the chain contributes also significantly to the permanent dipole
moment of this molecule. It is worth mentioning that the several more
complicated calculations we made using anisotropic correction factors,
such as those of [7], do not change significantly the results: the polar
angle is always large and the azimuthal angle such that the vector
always near the elongation axis.

8. THE MAGNETICALLY INDUCED BIREFRINGENCE.

( Knowledge of the complete a, € andx tensors now allows to calculate

*, Important molecular quantities such as the Kerr and Cotton-Mouton
-*' constants- Me shall limit ourselves to the latter, whose value is of
* prinary importance for quantitative analysis of magnetically induced

131
 118

birefringence (mib) data. The calculation of the former could be done in

a very similar way.
The molar Cotton-Moutton constant m C given by [11,13]:

(9)
15 kB7

with ,
= Tr (Ax0Ax0) (1°)
where ^a0 and Ax0 are the trace! ess molecular polar liability and
susceptibility tensors and NA is the Avogadro number. In our e.g. s
units, we have O0= a andXo/ kB = x / R - "here R is the gas constant.
Detailing the expression of the Trace, and using the values of the
tensor components in Tables 1 and 2, we obtain
15 3 2 1
m C (5CB) = (6.0 ± 1.O)XlO- Cm G- IiIOl- at T = 300 K. The error bar
corresponds to the six combinations of a. and x sets of Tables 3 and 4.
This value is comparable, but smaller, as expected, to the value for
biphenyl (6.99 in the same units) [U].
In the field of liquid crystals, a more experimental definition of the
Cotton-Moutton constant, noted CH, is used. One defines [14]:

An = CM x H2 (11)
where An is the experimental birefringence induced by the static
magnetic field H, and x is the wavelength of the light at which the
refractive indexes are measured. The relation between CM and m C is, for
a pure sample [14]:
3 K + 2)2 p 1

where n0 is the (isotropic) refractive index, p (g.cnr3) is the volumic

mass and M (g) is the molar mass. For solutions in apolar solvents, p/M ,
should be replaced by the molar concentration N (mol. cm'3) of (active)
molecules. For 5CB, the value (CM)0 of CM that would be measured if the
orientations of all molecules were uncorrelated is easily calculated
taking for m C the above value. At T NI = 308 K, on the isotropic side of
the clearing transition, we have n0(TNI) = 1.579 [6], p = 1.0092 g.cnr3 f.
\ [15]. For A = 6328 A, we obtain:
(CM)00= (3.6 ± 0. -13
'I Induced magnetic birefringence data in the isotropic phase of 5CB are '
reported in [8] and the results presented in terms of the experimental
constant CM versus reduced temperature. It is found that CM strongly
t increases as the clearing point is approached (from the high temperature t
Y side), and this phenomenon, which is common in all nematics, is V
;
-^ attributed to the increase of orientational correlations Detween the «j
î molecules. Phenomenological theory [16] predicts that CM should diverge \

132
 119

as (T - T1.)"1 where T0 is a temperature smaller but close to TNI. These

features are observed in 5CB, not too far from the transition [8].
From a molecular point of view, a more interesting quantity is the
ratio Nc= CM /(CM)0 which represents the average number of
orientationally correlated molecules. The isotropic phase, which is made
up of N molecules, may be pictured, in a first approximation, as the
juxtaposition of N / N0 independent entities, each of them being
constituted by N0 completely correlated molecules. This ratio has been
estimated as follows.
The values of CM are taken from figure 3 of [S]. On the other hand,
(CM)0 depends on temperature through n0, p and T. Assuming that n0- 1
is proportional to p, and using the volumetric data of [15], assuming a
constant expansion coefficient p = - dp/pdT of - 0.009 K"1, (CM)0 is
easily calculated at all temperatures. Figure 7 represents N0 versus
reduced temperature in a semi-log plot. It is observed that close to the
clearing point, Nc is of the order of 103 and decreases with increasing
temperature. The most important result for our purpose is however the
fact that at high temperature (presumably until the boiling point, which
cannot be much larger than that of cyanobiphenyl ~ 17O0C), a sizable
number of molecules, of the order of ten, remain correlated. This result
constitutes in itself a support for the solid-like picture of a liquid
at the molecular scale.

FIGURE 7

9. COMPARISON WITH ANALYSES OF THE LITERATURE

In all the papers from which the data have been taken [2-8], the
analyses have been systematically made, writing the equations in the
principal frame of the order tensor. Thus, all the measured quantities
are expressed as a sum of two terms, one proportional to Szz and the
other to Sxx- SYY . Then several more or less implicit attitudes have
been adopted. For example, in [2], it is argued that the biaxial term is
négligeable because Txx- Tyy is small. This implies that the principal
axis of S is assumed to be close to the para-axis of the biphenyl. Our
results in [1] shows that this is not quite true. It is also assumed
»•[ that the coefficient of Szz is independent of temperature. However, we
...J have seen that the orientation of the principal axes change with
temperature, and thus this assumption is not fully justified. Thus, the
power law which fits the temperature dependence xa is not exactly the
law followed by the order parameter.
;ji In [5], the emphasis is put on the comparison between the temperature
'I dependences of the magnetic susceptibility and of the refractive index.
Tne
'- finding that they are not the same is considered as an experimental

133
 •_^ ai
120

evidence that S10,- SYY is non zero in 5CB. However, in the analysis, it
is assumed that Oz is the common main principal axis of S, x and <*• We
have seen that this is not the case, and all conclusions can thus not be
taken too quantitatively. For example, it is argued in [5] that the
magnetic susceptibility is more sensitive to the biaxiality of the order
than the refractive index. This argument cannot be general and depends
on the particular orientations of the three tensors S, x and a. For the
case of 5CB, this is illustrated in Figure 8 where are represented the
(reduced) temperature dependences of the three quantities SZ2, xa and
Sn, where the latter have been scaled by a multiplication factor S0 such
that all the curves are comparable).

FIGURE 8

FIGURE 9

Although the three curves are close, they are not the same. The curve
for the refractive index is indeed closer to that of Szz. This result
may appear as a support to the argument of [5]. However since in [5], it
is assumed that the principal axis of S is the para-axis, the argument
holds in fact for T22. We have thus repeated the above analysis with T22.
, and the results are shown in figure 9, which is similar to 8, but with
Szz replaced by T22 , and the scaling factors S0 replaced by other
factors T0. It is observed that the situation is different. Now, the
refractive index follows Tzz at high temperature only. But globally, the
magnetic susceptibility represents better the temperature dependence
over the whole temperature range.
Finally, in [7], the analysis of the permittivity data is made
assuming uniaxial symmetry around the same axis for all the tensors. The
results are used to determine the single order parameter versus
temperature. Their results are significantly different from ours. In
addition, these results emphasize the importance of the internal field
correction factors. In this respect, it is remarkable that the more
sophisticated are the corrections, the less reliable appear the results.
Thus, this method does not seem the best one to determine order
parameters. Conversely, we have shown that if the order ami polarization
tensors are imposed, the results concerning the dipole moment are rather
stable with respect to these correction"-.

:y 10. RELATION BETWEEN THEORETICAL AND EXPERIMENTAL ORDER PARAMETERS.

This discussion shows that the question of which physical quantity is

 121

better to represent the macroscopic "order parameter" of a nematic phase

is in fact not a good one, or at least it is badly stated. A priori,
all second rank tensorial physical quantity (the symmetry of the phase
should be respected) which vanish at the transition are equally valid.
Thus, all fourty magnetic interactions considered in [1], as well as the
three quantities in this paper can be chovn. However, the trouble comes
when one tries to compare the temperature dependences with theoretical
predictions. We have seen in this study that these dependences are not
the same. This is not surprising since the measured quantities are
linear combinations of five (statistical) temperature dependent
parameters, namely the elements of the order tensor, whose coefficients
are (molecular) parameters which strongly depend on details of the
molecular structure and possibly also of temperature if there is an
associated change in the conformation. From the theoretical side, two
complementary descriptions of the nematic phase are currently used: the
phenomenological, and the molecular mean-field, theories [16,17].
The order parameter S of the phenomenological theories is a very
general concept. The (Landau) free energy of the system is developed in
powers of S . The simplest version of the theory predicts that, within
the nematic range, S varies with temperarure as (T - T-,)1'2 where T0 is
a temperature slightly higher than the clearing temperature T NI . The
value 1/2 for the exponent comes from the assumption that the
coefficient A of the term proportional to the square of the order
parameter in the expansion of the free energy, vanishes at the
transition as T - Ta . A universal behaviour is thus predicted for all
the possible order parameters that can be defined. This is not observed:
although power laws are indeed observed, the exponents are different
from 1/2, and depend on the particular quantity chosen to represent S
[5]. This result thus appears as a limitation of this simple
description. Improvements of the theory are currently made to explain
why the exponent is different from 1/2 [18].
The order parameters of mean-field theories on the other hand, depend
on an essential way on the particular form assumed for the anisotropic
intermolecular potential. These potentials are always chosen
considerably simpler than the actual ones ciue to the complexity of the
real molecules. In fact, a "phase adapted' or an "experiment adapted"
choice of the potential is generally made, in that sense that the
virtues of the phase, or of the measured quantities, are attributed to
the molecules (more precisely to the interparticle potential). In the
former theories of Onsager, Flory, Maier and Saupe, and in their
generalizations, the molecules are assumed to be perfect cylinders,
because the phase was uniaxial. When it was realized that the order (of
the molecule and/or of the phase) could possibly be biaxial (an
experimental consequence, due to the second rank tensorial character of
the measured quantities), the shape was changed to parallelipipeds or
. elliptical cylinders, and the theory could predict two order parameters
1 and their temperature dependences. But actual molecules are (very) far
from all these simplified models, and it is why the universality of the
 122

temperature dependence of the order parameter predicted by e.g. the

Maier-Saupe theory, is not observed. In this sense, the value of these
models is only qualitative [17]. Furthermore they cannot predict, for
example, the values of the melting and/or clearing transition
temperatures, which are of primary importance for practical uses. In our
opinion, this situation will persist as long as a realistic modelling
(e.g. by taking into account the experimental information obtained in
the solid phase) is not introduced in the theories.
A review on the more recent developments of the two complementary
theoretical descriptions of the nematic phase can be found in [18].

11. CONCLUDING REMARKS.

In this second paper, the results of the analysis of the NMR data in
terms of the single conformation model, performed in [1], have been
applied to analyze data other than NMR. We have found that not only no
contradiction with known results have been found, but also rather
specific information concerning molecular tensors have been deduced. In
themselves, these results may be considered as new molecular
information, or conversely as a possible way to test the single
conformation model if this new information could be obtained in an
independent way. A possibility is quantum chemistry calculations.
Although considerable progress has been made in this field in recent
years, this method may not be quantitatively quite reliable when used to
find the structure and conformation corresponding to the minimum energy
of relatively large molecules such as 5CB, with external constraints due
to interactions with neighboring molecules. However, here, the problem
is different: the structure and conformation are known, and it would
then be possible to calculate completely all the tensors, and make
comparison with the experimentally predicted tensors. Thus a number of
predictions emerge from this study, which could be used as a further
test of the model.
We would end this paper with some philosophical considerations
concerning the single conformation model used in this and previous
works. The main interest of a simple model is not that it works
everywhere, but that there is some place where it does not. In all our
previous papers, cited in [1], in which NMR data of a number of nonrigid
molecules have been analyzed in terms of the single conformation model,
we have found that in all cases, it worked. Very recently, another
important example has been reported with succinic acid: seven
independent data could be analyzed with only four adjustable parameters
[19]. The case of 5CB treated here is intrinsically less stringent.
However, we have seen that the model in its strict sense fails for this
molecule. In our opinion, this failure is in itself the most important
result of all this study, since it is the signature of a new phenomenon:
1 the end-chain disorder. However, the perfect model now appears to be not
so perfect. The perfection of the internal motions (i.e. the fact that
 123

they are only symmetry operations) is broken at the extremity of the

chain. But a perfect thing can only be replaced by another perfect thing
[2O]. Thus, the perfection of the single conformation model should be
tranferred at a level other than the molecule. In line with the
solid-like picture of a liquid, picture which begins to be invoked also
for non-organic liquids [21], we conjecture that this level is that of
the unit cell. More precisely, the symmetry of the internal motions is
recovered at the scale of the unit cell. This aspect will be developped
in a forthcoming paper.

12. ACKNOWLEDGMENTS

The authors are indebted to Doctors G.Maret and B. Malraison for

useful discussions about theoretical and experimental aspects of induced
birefringence. This work was partially supported by NSF/CNRS Grant INT -
871501/88 - 920156.

''I
Vi

f
 124

REFERENCES AND FOOTNOTES

[1] Gérard H, Galland D. and Volino F., Th*- -jreceeding paper of this
Issue.
[2]Buka A. and de Jeu W.M., J.Phys.France 43 {193?) 3-'.
[3]Sherel1 P.L. and Crellin D.A., J.Phys France 40 (la/9) C3-211.
[4]Frisken B.J. Carolan J.F., Palffy-Muhoray P., Perenboom
J.A.A.J. and Bates G.S., MoI.Cryst.Liq.Cryst.Letters 3 (1986) 57.
[5]Bunning J.D., Crellin D.A. and Faber T.E., Liquid Crystals I
(1986) 37.
[6]Horn R.G., O.Phys.France 39 (1979) 105.
[7]Dunmur D.A, Manterfield M.R., Miller W.H. and Dunleavy J.K.,
MoI.Cryst.Liq.Cryst. 45 (1978) 127.
[8]Filippini J.C., Poggi Y. and Maret G.,"Colloques Internationaux
C.N.R.S." (France) N°242 - Physique sous Champs Magnétiques Intenses
(1974).
[9]Krishnan K.S., Guha B.C. and Banerjee S. Phil.Trans.Roy.Soc. A231
(1933) 235.
[10]ref.[7] of (I]
[ll]The Late Flory P.J. and Navard P., J.Chem.Soc., Faraday
Trans.1, 82 (1986) 3381.
[12]Coles H.J. and Jennings B.R., MoI.Phys. 36 (1978) 1661.
[13]Cheng C.L., Murphy D.S.N. and Ritchie G.L.D, J.Chem.Soc., Faraday
Trans. 2, 10 (1972) 1679.
[HJMaret G. and Dransfeld K. in "Topics in Applied Physics" 57,
Herlach F. Editor; Springer-Verlag Berlin Heidelberg (1985)p. 143.
[15]Dunmur D.A and Miller W.H., J.Phys.France, Colloques C3-40
(1979) C3-141.
[16]de Gennes P.G."The Physics of Liquid Crystals", Clarendon
Press (1974).
[17]Vertogen G and de Jeu W.H. "Thermotropic Liquid Crystals,
Fundamentals", Springer Series in Chemical Physics 45 (1988).
[18]Prost J."Theory of Liquid Crystals", Clarendon Press, in print
[19]Chidichimo G., Formoso P., Golemme A. and Imbardelli D.
(preprint: personal communication).
[20]Feynman R.P."The Character of Physical Law", Messenger
Lectures, The M.I.T. Press, Sixth Edition (1975).
[21]Vlasov A.D and Rez J.S., Cryst.Res.Techno!. 26 (1991) 611.
Vl
 ,. ._iJ***S

125

TABLES

Table 1^Components of the magnetic susceptibility tensor of 5CB,

in the molecular frame defined in [1], obtained from analysis of
the data published "in the references mentioned. Units:

Reference Xx,

[2] - 141.2 - 136.6 - 224.1 - 26.3 - 1.5 - 4.4

[3,5] - 137.0 - 138.7 - 226.2 - 26.3 - 4.1 - 3.4
[4] - 137.1 - 138.7 - 226.1 - 26.3 - 2.0 - 2.4

Table 2._ Components ot the optical polarizability tensor of 5CB

(A = 6328 A), in the molecular frame defined in [1], obtained from
analysis of the data of [5], for two values of the internal
field correction factor C. Units: A 3 .

Factor C «zx

1.00 46.0 29.9 23.7 3.5 2.3 - 1.9

1.24 43.8 32.5 23.3 6.5 2.2 - 0.3

I*
•I
,fv
\
 126

'8
35

30

25

20

15

10

-20 -16 -12

Fig.l Anisotropy of magnetic susceptibility xa(units of 10~6cm3mol"1 ) of

nematic 5GB versus reduced temperature. The points are experimental.
Losanges: data from [4]. Triangles: data from [3.5], Circles: data from
[2]. The lines join calculated values.

ir

1,
 •Si.-

127

L
X Y

X Y

5?*
J Fig.2 Stereoscopic projections along principal axes OX (upper) and OY
! (lower) of the magnetic susceptibility tensor x. of the conformation of
5CB predicted by the single conformation model. The principal axis OZ is
vertical. The temperature corresponds to the middle of the neraatic
phase. The cyano group and end-chain disorder are not shown for
simplicity. See [1] for details.
 128

'n

.3

.2

.1

-20 -16 -12 -8

s»
Fig. 3 Anisotropy of the optical refractive index Zn at 6328 A of nematic
5CB versus reduced temperature. The points are the data of [5]. The
3\ lines join calculated values for the two values of the internal field
..* tactor. The two calculated lines are not discernable on this scale.
 129

L
X Y

J
X Y

Fig-4 Sa016 as figure 2, for the optical polarizability tensor a.

 130

1.5

1.2

.9

.6

.3

-20 -16 -12 -8

'
l Fxg.5 Anisotropy of the electrical permittivity ££in the audio-frequency
range (1592 H2) of nematic 5CB versus reduced temperature. The points
are the data of [7]. The line joins calculated points. Note that the
data extends over a reduced temperature range.

1
 131

Fig. 6 Sane as figure 2, for the tensor A. Note that this tensor is
axlally synaetric around the direction of the permanent dipole moment u.
= OZ. The two principal axes OX and OY have been (arbitrarily) chosen
such that the para-axis of the biphenyl either makes the maximum angle
with the dipole moment (upper), or proj ects along the dipole moment
(lower). The end-chain disorder is shown (as in figure 12 of [I]).

,1
 132

1000:

100::
A
«A

10::

20 40 60 80 100
''I Fig.7 Reduced Cotton-Mouton constant NC- CM/(CM)Qof isotropic 5CB versus
reduced temperature. The values of CM are taken from [8]. The values of
(CM)0 are calculated as explained in the text. The triangles and circles
delimitate the systemat. uncertainty associated with the order tensors
y and a quoted in Tables 1 and 2. Note the relatively large value of NC
~ 10, at high temperature.
t.

']

149
 133

.75

.65

.55

.45 n

.35
-20 -16 -12

'I Fig.8 Comparison between the (reduced) temperature dependences of Szz

«1 (triangles), xa (broken line) and ^ (continuous line).
 ,-•«î

134

.7

.6 .. »22

.5

„4 'n

.3
-20 -16 -12 -4 O

s*

Fig.9 Same as figure 8, with Szz replaced by T

i#
\
 135

n C Structure et ordre orientationnel de polymères nématloues

1inéa1resr24]
Dans cette étude, le modèle à conformation unique a été employé
afin de calculer les spectres proton de l'unité mésogène de polymères
nématiques (les [poly [oxy(3-méthyl-l,4-phénylène) azoxy(2-méthyl-
1,4-phénylène) oxy(a,u-dioxy- a,w-alkanediyl)]) notésME9Sn et d'en
tirer des informations tant sur là structure de cette unité que sur
l'ordre y étant associé.
L'unité mésogène considérée est constituée du groupe azoxybenzène
comprenant deux groupes methyls substitués sur les groupements phényl.
La deutération de l'espaceur flexible (CH2)n permet de réduire la
complexité du spectre de RHN du proton de ces molécules (chaque
monomère de ME9Sn comprend 12 + 2n protons, la prise en compte des
seules unités mésogènes permettant de revenir à un système de 12
protons pouvant être considéré comme magnétiquement isolé). Cependant,
l'utilisation de polymères partiellement deutérés ne résoud pas les
autres problèmes inhérents aux polymères, qui sont l'élargissement des
formes de raie dû aux modes élastiques lents et la polydispersité
(hétérogénéité des échantillons due à la ségrégation entre chaînes de
1ongueurs di fférentes).
Un moyen d'obtenir une information de départ sur les polymères de
la famille ME9Sn est de partir de la molécule de 4,4'-diacetoxy-
2,2'-dimethylazoxybenzène (Ac9Ac) que l'on considère en tant que
modèle pour l'unité mésogène de ces polymères. Cette molécule seule
présente un faible caractère mésogène et ne présente pas de phase
nématique, mais on peut en obtenir en la mettant en solution dans des
solvants nématiques tel le para-azoxyanisole (PAA). Cette "petite"
molécule, échappant aux deux problèmes, évoqués plus haut, liés à
l'étude par RMN des polymères, pose néammoins deux autres problèmes
qui sont i) la présence des deux methyls terminaux et ii) le temps d'
homogénisation important pour les mélanges Ac9Ac/PAAdl4. La première
de ces difficultés peut être contournée en utilisant de T Ac9Acd6,
deutéré sur les groupements methyls terminaux, la seconde avec de la
patience et en travaillant dans la biphase N+I du mélange.
Dans le chapitre précédent, le modèle à conformation unique a été
utilisé afin de calculer des interactions magnétiques déterminées
grâce à diverses méthodes RMN plus ou moins sophistiquées. De telles
données n'étant pas disponibles pour les polymères considérés ni pour
la molécule modèle, le nombre assez élevé de spins (douze) considérés
posant en outre des problèmes d'attribution, la méthode de travail
utilisée ici, plus complexe mais finalement plus riche en information
a été la suivante, composée de trois étapes: on suppose une
conformation pour la partie mésogène comprenant douze spins 1/2, ce
qui permet de calculer les interactions magnétiques à partir du modèle
à conformation unique (en supposant divers mouvements internes) puis
de calculer un spectre que l'on peut comparer aux spectres
 •-, .«JE

136

expérimentaux disponibles.
Cette méthode de travail a déjà été appliquée avec succès dans la
cas du PAAdG dont les résultats vont constituer une base pour cette
étude, cette molécule étant constituée d'un groupe semblable à l'unité
mésogène des ME9Sn mais dont les phényls ne sont pas substitués par
des groupements methyls, et de deux groupes methyls terminaux
(deutérés dans le PAAd6). La différence essentielle entre les
polymères considérés, et leur molécule modèle, et le PAAd6, outre le
nombre plus élevé de spins, réside dans la présence de methyls sur les
phényls qui implique que les rotations de ir de ces derniers ne
correspondent plus à des opérations de symétrie, et ne seront donc pas
prises en compte dans le calcul des interactions magnétiques.
En supposant que l'ordre est uni axial (le tenseur d'ordre pouvant
alors être représenté par seulement trois paramètres qui sont les
trois angles d'Euler (le premier d'entre eux étant arbitraire)
décrivant l'orientation de son repère principal par rapport à un
réfërentiel moléculaire et un paramètre d'ordre unique S0=S2Z), on
peut obtenir des spectres reproduisant de manière satisfaisante les
spectres expérimentaux de VAc9Acd6 en solution dans le PAAdH et de
deux polymères ME9Sn deutérés sur leur espaceur flexible, TAZA9dl4 (n
= 7) et le DDA9d20 (n=10). Ces simulations ne sont pas parfaites, ce
qui se comprend bien dans le cas des deux polymères où l'on n'a tenu
compte ni des fluctuations élastiques lentes ni du phénomène de
polydispersité. Dans le cas de TAc9Acd6, exempt des deux
inconvénients précédents, cette imperfection peut s'expliquer par le
fait que i) les approximations faites dans la description de l'unité
mésogène (utilisation de deux valeurs différentes d'anisotropie de
déplacement chimique, pour les protons des phényls ou des methyls, au
lieu de huit dans l'absolu, hypothèse de la planitude du groupe azoxy
CNNC) ne sont pas forcément adéquates, ii) une légère incertitude
existe toujours sur les formes de raie expérimentales et iii) la
deutération partielle peut présenter des imperfections, la présence de
protons dans les methyls terminaux pouvant contribuer au signal
détecté. Ces trois réserves s'appliquent bien sûr également aux
spectres de polymères.
Malgré ces approximations, le calcul rigoureux du spectre proton
de l'unité mésogène de ces trois molécules a permis d'obtenir nombre
d'informations tant sur leur conformation que sur Tordre au sein de
leur phase nématique. En ce qui concerne ce dernier, les deux plans
principaux XOZ et YOZ du tenseur d'ordre correspondent
approximativement au deux plans bissecteurs des phényls (dans le cas
où l'on suppose une biaxialité de ce tenseur permettant de distinguer
OX et OY), et l'unité mésogène apparaît mieux alignée selon OZ dans
les polymères que dans TAc9Acd6, Cependant, cette étude ne permet pas
I de déterminer si le tenseur d'ordre est biaxial, d'aussi bons
résultats étant obtenus avec ou sans biaxialité. D'un point de vue
\ conformât!onnel, on retrouve, comme dans le PAA, une distortion
importante entre les orientations des deux phényls, qui se traduit par
 137 O

la dissymétrie de leur contribution au spectre. Enfin, on a obtenu une

estimation de la différence des projections des tenseurs de
déplacement chimique des protons du phényl ou du méthyl selon OZ,
ainsi qu'une preuve en faveur de l'absence de rotation de TT des
phényls autour de leur axe para.
Les résultats obtenus pour l'AZA9d!4, un des deux polymères
considérés, ont servi de base à l'étude de la variation des spectres
de RHN du proton de ce polymère en fonction de Tangle 0 entre le
vecteur directeur n^ de sa phase nématique et le champ statique fÇ de
l'expérience de RHN. Cette étude est exposée dans l'article consacré à
la théorie de l'influence de modes visco-élastiques sur les formes de
raie RMN, appliquée pour la première fois, et avec succès, à ce
polymère nématique.

5
1

i
 139

Reprinted from Mtcrotnoleeules, 1992. 25. .

Copyright © 1992 by the American Chemical Society and reprinted by permission of the copynght owner.

Exact Calculation of the NMR Spectrum of a Twelve Spin 1/2

System: Application to the Structure of the Aromatic Core of a
Main-Chain Nematic Polymer
D. Galland, H. Gerard, J. A. Ratio,» and F. Volino"
C-E.A.-C-EJV. Grenoble. DRFMC/Service d'Etudes des Systèmes et Architectures
Moléculaires, Groupe Physico-Chimie Moléculaire, 8SX, 38041 Grenoble Cedex, France

J. B. Ferreira
Centra de Fùica da Matiria Condensada, Z, Ao. Prof. Cornu Pinto, 1S99 Lisbon Codex,
Portugal
Received August 18. 1991; Reused Manuscript Received April 30, 1992

ABSTRACT: The anisotropic proton NMR spectrum of the 2^-dimethyIazoxybeiuene moiety, a twelve
spin Vi system, is calculated eiactly and compared to the experimental spectra of (i) 4,4'-diacetoxy-2,2'-
dunethylazoxyDenzene deuterated on the methyl groups (Ac9Acd6) and dissolved in the nematic phase of
perdeutented p-azoxyanisolt (PAAdU) and (U) two spacer deutented main-chain thennotropic nematic
polymet^lpolytoiy<3-methyl-l,4-pheDylene)aioiy(2-methyl-l,4-phenylene)oxy(a^-dioiy-a^-»lkanediyl)]
(MESSn, with n = 7 and 10 (AZA9dl4 and DDASdZQ)] in their bulk nematic phase. Modeling in terms of
the single conformation model allows satisfactory fits of the expérimental spectra. The structure and ori-
entational features of this mesogenic moiety in AcSAc and in the polymers are found to be very similar.
Similarities and differences with the parent moiety atoxybenzene (the aromatic core of PAA) are pointed
out. The similarities lie mainly in very similar distortions of the backbone, constituted by the «zoiy group
and the two para axes of the rings, and in the orientation of the principal axis. The différences lie in the
absence of r-flips of the rings (such flips exist in PAA and are uneorrelated) and in larger values of the
dihedral angles between the three rigid moieties. The simulations permit the determination of the
proportionality constant between the main splitting of the spectra and the uniaxial order parameter. Some
weak information on the anisotropic components of the chemical shift tensors of the phenyl and methyl
protons and on the biaxiality of the order is also obtained.

I. Introduction dispersity (heterogeneity of macroscopic samples due to

As a continuation of our work1-2 on molecular aspects
in nematic main-chain polymers of the form (RF)n where
R is a mesogenic unit and F a flexible spacer (CHJn, we
have undertaken the detailed study of the proton NMR
spectrum of thfimesogenic unit of polymers MESSn^4 with
chemical formula
A recent review of the properties of these polymers, with
numerous references, is made in ref E.
The mesogenic unit is the azoxybenzene moiety where
two methyl groups are substituted on the phenyl rings as
indicated. This study is aimed at obtaining information
on the structure and conformation of this unit as well as
itaorientationalorder. Forthispurpose.exactsimulation
of the proton NMR spectrum is one of the most powerful
methods provided the experimental spectrum is suffi-
ciently well resolved (that is, it exhibits sufficient signif-
icant features) and the number of spins is not too large.
In the case of MESSn polymers, these two conditions are
not fulfilled because (i) the number of spins per repeat
unit is very large (12 •+• 2n) and (u) there are broadening
effectad-jetoslowmotions(alowelasticmodese)andpoly-
segregation by chain lengths7).
The first complication can be practically eliminated by
deuteration of the spacers, thus reducing the number of
protons to only 12. Each polymer can now be considered
as the juxtaposition of z mesogens without magnetic
interaction, because two consecutive mesogens are far apart
from each other. Moreover, the linkage carboxylic group
is magnetically inactive and relatively bulky, so that the
broadening effect on the proton spectrum due to dipole
interaction with the spacer deuterons is also very weak.
The broadening effects due to the polymeric nature remain
however, which blurr details of the spectra.
The broadening effect due to viscosity can be reduced
by using short polymers. However, end-chain effects
become more important and increase the degree of
heterogeneity: the width of the nematic phis isotropic
(N-U) biphase indeed increases forME9Sn polymers when
the average moleci. ji mass decreases.5-*
The molecule4,4'-diacetoiy-2^'-dimethylazoxy benzene,
hereafter labeled Ac9Ac, with chemical formula
CH1COO
CH3
may be considered as a model for the meaogenic unit of
ME9Sn polymers.
Mixtures of Ad)Ac with perdeuterated p-azoxyanisole

i 'Towhom correspondence ibould be «ddieMed. Member of the

(PAAdU) have been studied in detail, and the results are
Centre Nitionil de Ia Recherche Scientifique (CUJLS.). France. described in ref 9. Bulk AcSAc is not nematic. The
* On fane from the PoJymerPromm, Department of Chemistry. molecule has a poor mesogenic character, the virtual
Univenity of Lowell, Lowell, MA 01854. clearing point being about 220 K,9 compared to 409 K for

0024-9297/92/r225-4519S03.00/0 C1992 American Chemical Society

Lr:
 140

4520 Gallandetal. Mocromoleeuies, Vol. 25, No. 18. J992

PAA. Th* consequence of this situation is the existence

of a broad N+Ibiphasicrangeinthet'AAdU/AcSAcphase
diagram and of very long times to obtain homogeneous (a)
samples in NMR tubes, in which stirring is not possible.7
The equilibrium properties of several Ac9Ac/PAAdl4
mixtures are described in ref 9, and the kinetics of
hotnogenizaUon of these mixtures as well as of mixtures
of ME9Sn polymers with PAAdM is described in ref 7.
Although the use of a small molecule in a low molecular
massnematic allows one to getridof the broadening effect
due to polydispersity and viscosity! two problems remain:
(i) the existence of the two terminal methyl groups in the
AcSAcmolecule and (ii) the difficulty of obtainingA?9Ac/
PAAdl4 mixtures with sufficiently uniform composition
over the sample volume, ~0.1 cm3. The first difficulty
has been solved by synthesizing Ac9Acd6, deuterated on
the terminal methyl groups, but the second difficulty can
onlybesolvedbypatience. ID fact, the most homogeneous
nematic state is obtained when the sample is in the N+I
biphase. The external part of the spectrum, which cor-
responds to the nematic component, is well resolved, but,
clearly, the central part is completely hidden by the sharp
signal of the isotropic component.
Figure 1 illustrates the effects described. Figure Ia is
the proton spectrum of spacer-deuterated ME9S 10 poly-
mer with z ~ 5.6 (hereafter called DDA9d20 polymer),
andFigurelbis thatof spacer-deuterated ME9S7 polymer
with i ~ 10 (hereafter called AZA9dl4 polymer), both in
their nematic phase. These two polymers are the same as
those used in ref 2. Figure Ic is the spectrum of Ac9Acd6
in solution (15% w/w) in PAAdU at =90 0C in the pure
nematic phase of the mixture, and Figure Id is the
spectrum of the same sample at a higher temperature,
=116 0C, in the N+I biphase. It is observed that all these
spectra are similar, although some small differences are
apparent, in particular in the central part of the spectra.
They are not symmetrical with respect to the center,
showing that the various spins (essentially the phenyl and
methyl protons) do not have the same chffmicrf shifts. Figure 1. Experimental (upper) and simulated (lower) proton
Spectrum Id appears to be the best resolved spectrum, NMR spectra of DDA9420 polymer (a) and AZASdU polymer
Hlthnueh the central part is completely hiddenby the sharp (b) in their bulk nematic phase «id of Ac9Acd6 dissolved in
isotropic component Comparison between all spectra PAAdU at =90 °C in the pure nematic phase (c) and at »116
0
suggests that the central sharp peak on the spectrum of C in the nematic plus isotropic biphase (d). The uniaxial order
parameter So deduced from eq 1 and the relative chemical shift
the polymers (Figure la,b) is probably due to residual UkU (in ppm) used in the simulations are respectively 0.77 and 3.8
(a), 0.485 and 3.4 (b), 0.50 and 3.4 (c), and 0.33 and 3.6 (d). Note
of the N-I-I biphase which exists at higher temperature. that the dipolar interactions used for spectra b and d are the
This observation substantiates the effect of heterogeneity same as those used for spectra a and c given in the Appendix,
of composition on the proton line shapes. This effect is respectively, properly scaled by So.
better seen on the deuterium line shapes of the polymers
(cf. the spectra published inrers1 and 2), and this aspect should be split into two parts. The first part would be to
wDlbediseussedindetailebewhere. The present purpose measure the dipolar interactions from the NMR exper-
is to obtain information on the structure and orientational iments, and the second part would be to interpret these
order of the mesogenic unit inAc9Acd6 and in the MESSn data in terms of molecular structure, conformation, and
polymers from exact simulation or the corresponding motions. The first part of this task becomes very difficult

i twelve spin Vz system.
2. Determination of Molecular Structure from
Simulation of NMR Spectra
Simulation of the (single quantum or multiquantum)
proton NMR spectrum ofamolecule containing Nprotons
in & fluid anisotropic medium requires knowledge of the
NlN - 1)/2 dipolar interactions and of the N chemical
shiftswithzespecttoanarbitraryreference. IndiredJ-J
couplings are also of interest, but they are usually small
and they will be neglected bare. The dipolar interactions
as soon as the number of spins becomes huge (say larger
than eiçht spins) and requires sophisticated multipulse
methods and/or selective or random partial deuteration.
This approach has been applied recently to n-hexane (a
14-proton system) in solution in a nematicsolvent.10 Even
here, however, the complete unequivocal assignment of
the interactions is impossible directly, and one his to rely
on the use of models. This shows that, in practice, the two
parts of the problem cannot be solved independently when
the number of spins is large, as is the case here.
The other.more pragmatic, waytoapproach the problem
is to assume a structure for Ae molecule, choose a model

i
are directly functions of the structure of the molecule and for the motions and order, calculate the dipole interactions,
of the (anisotropic) motions of th; spins. The fh»mica1 simulate the spectrum, and compare with the experimental
shiftsarealsofunctionsofthelatter. Ideally.the problem spectrum. This trial and error method turns out to be
 141
Maeromolecules. Vol. 25, No. 18.1992
9.
A conformations are described by the same order tensor.
(n)
figure 2. Sketch of the 2^-dnnethyIazGiybenzene moiety
showing (i) the main bond angles, bond distances, and molecular
frames used in the study and (U) the labeling of the 12 proton
spins.
very successful when sufficient reliable information is pri-
orly known, in particular structural information. This
method is used by several groups.11-12 It is important,
ho we ver, that the calculations are performed exactly.1*43
For the particular case of nematic PAA, the exact
simulation of the proton spectrum of PAA deuterated on
the methyl groups—an eight-proton system—using results
obtained from deuterium NMR data has unambiguously
demonstrated that the structure of the core is very close
to that in the solid phase, thatthe two phenylringsperform
uncorrelated i-flips, and that the orientational order is
biaxial.13"15
The present problem may be considered as a general-
ization of the PAAd6 problem in the sense that the core
is the same, the only difference being the substitution of
protons in positions Ig on the rings by two methyl groups.
3. Modeling the 2£'-Dimethylazoxybenzene
Aromatic Core
This moiety is constituted by the succession of three
rigid moieties, namely the methyl-substituted phenyl ring
B, the CNNOC azoxy group, and the methyl-substituted
phenyl ring A. linked by two single covalent bonds (Figure
2). InPAA, the slignmentsofthetworingsaredifferent15
so that one might expect that the situation is the same
here. A first indication of this asymmetry is given by the
deuterium spectrum of Ac9Acd6 in PAAdU, which shows
that the alignment of the two terminal methyl groups is not
the same. This nonequivalence will become obvious when
attempting to simulate the proton spectra.
An important aspect is that of the internal motions.
Clearly, methyl group rotation, or rather jumps among
three equivalent positions, are highly probable. The
question of the ring flips about their para axis, which exists
in PAA (more generally, in all molecules with normal or
symmetricallysubstitutedphenylrings: toourknowledge,
there is no known counterexample in fluid phases), is
'! different here due to breaking of the Ca, symmetry of the
rings bythe presence of the methyl groups in the2position
(and in the T position). In several papers,13-17 we have
argued that, due to the similarity of the local molecular
packing in solid and liquid phases of the same substance
(solidlike picture of a liquid17), the only large-amplitude
internal motions (Le., motions which produce significant
averaging of the magnetic interactions) allowed are sym-
metry operations. We shall thus assume here that the
t T-flips of the methyl-substituted pheny ! rings are forbid-
den and thatthe only internal motions within the aromatic
} core are the methyl group rotations and exchange with
the mirror image conformation. The conformations ob-
NMR Spectrum of a Twelve Spin V2 System 4521
tamed after these symmetry operations are identical (or
very similar), and the orientational orders of these
This allows the problem of the magnetic interactions in
such nonrigid molecules to be treated in a completely rigid
molecular geometry, inside which the spins are allowed to
exchange via the symmetry operations. Since all the
conformations are known if one of them is known, this
model is called the single-conformation model (the name
single-order tensor model would however be more appro-
priate). These ideas are discussed in detail in ref 17, but
also m refs 14-16.
The calculation of the various dipole-dipole interactions,
whose expressions are given elsewhere,16 have been per-
formed as fc'iu- js. A frame is attached to each fragment,
as depicted in Figure 2. Frame OiiyiZi (frame B) is
attached to ring B, with Ozi along the para axis and Oxi
in the plane of the ring and directed toward the side where
is located the methyl group. A similar convention is made
for ring A, with index 1 replaced by index 4 (frame A). The
rings are assumed to be perfect hexagons, with CC = 1.39
A and CH = 1.085 A. The angles CCH are fixed at 120°.
With these values, the distance between ortho protons is
2.475 A. A frame is also attached to the methyl groups
with Oz along the threefold axis and the Ozz plane
containing one methyl proton (by convention, proton 3
formethylB and proton 12 for methyl A). The orientation
of the methyl frame with respect to the ring frame is
described by three Euler angles. The first one is clearly
zero, the second is assumed to be clo^e to 60° for ring B
and close to 120° for ring A (they are fixed at these values
in a latter stage of the simulation; see below), and the
third angle «vn describes the orientation of the methyl group
(«5m = O when proton 3 (or 12) is in the ring plane). The
CC,«t distance was chosen very close to 1.5 A, according
tothe structure of o-and p-polvureau and of vicinal poly-
ketones,19 where this distance is found to lie between 1.495
and 1.525 A, depending on the particular methyl group
considered. The CH distance in the methyl groups was
fixed at 1.0946 A and the CC0nH angle at 109.47° as in
ethoxybenzene.16 In polyketones19 it is in addition ob-
served that the orientations of the several methyl groups
with respect to the ring planes do not have any particular
symmetry, implying that the two angles v>m are not zero
or T. They have been taken as adjustable parameters in
our problem.
The two rings are linked by the CNNOC azoxy moiety.
Its structure is assumed to be planar, with bond angles
and distances identical to thoseofPAA in the solid phase.20
(This structure was also cbaseo io simulate the 1H NMR
spectrum of PAAd6.u) Ti:a values are given in Figure 2.
Finally, the last problem concerns the relative orien-
tations of the two rings with respect to the azoxy group.
In PAA, the striking feature is that in the solid phase20
(and probably also in the nematic phase14) the two para
axes of the rings are neither parallel nor coplanar, and
significative distortions exist, corresponding to misori-
entations of the outgoing CN bonds of several degrees
with respect to the corresponding para axes. We have
thus introduced two frames, Ox2yiZ2 (frame 0} and 0x3X323
(frame a), with Oz along the CN bonds and Ox in the
planeof the azoxy group and oriented as indicated in Figure
2.
The motional parameters are reduced to a minimum.
There is no free parameter for the internal motions. The
external motions are described by the five elements of the
(single) order tensor. For simplicity, it will be assumed
that the order is uniaxial (the biaxiality is probably weak,

142
4522 Gallandetal.
2
in any case much weaker than in PAA ) and the remaining
single-order parameter Sa is just a scaling factor. The
orientation of the principal frame with respect to the B
frame is described by three Euler angles, Ui, V1, and Wi.
The first angle is undetermined (uniaiiality), while the
other two angles represent the polar and azunuthal angles
of the principal aiis OZ in the B frame. These two angles
correspond to the only effective motional parameters of
the problem.
The geometrical parameters are the ten unknown angles
mentioned below. The orientation of the P frame with
respect to B is described by Euler angles Ut, V2, and W2, .
the orientation offramea with respect toframeftby Euler
angles O. -65J" + 68\2° » 2.9", and O (the geometry of the
azozy group is assumed to be that of PAA), and, finally,
the orientation of frame A with respect to frame a by
Euler angles I/a. V3, and W3. We recall the two angles for
themethylgroups, namely the orientations of the threefold
axis in the ring planes, close to 60 (120)°, and the angles
Vb-
In total, there are 12 (in practice, only ten; see below)
free parameters to be determined from the simulation.
This seems a priori r. very difficult task to determine
immnhipiimJy all Oifgt pyainetenifmm a singlp aportnim
such as those of Figure 1. However, it will be seen that
the features of these spectra are so particular that
practically only a structure (very) close to those that will
be proposed can account for these features in the frame-
work of the single-conformation model.
4. Simulation Procedure
As mentioned in previous work,13 eiact simulation of
proton spectra seems to be necessary to obtain very good
(and thus unequivocal) results. The reason is that, in such
molecules, it is not possible to split properly the total spin
system into subsets with smaller numbers of spins (as
required by the approximate methods11-13), because there
are always interactions between spins of two subsets which
arelarge. However.suchasplitmayhelpinunderstanding
the origin of some spectral features, as shown below.
spin Vj system considered here is an improved version of
the algorithm used previously in thecaseofPAAdG.13 The
improvement lies in two facts: it can be used for any
number of spins (the limitation is only the memory size
of the computer) and the chemical shifts of all spins may
be different. The full spectrum for twelve spins was
calculated using a CRAY2 computer, while intermediate
six-spin spectra were calculated with a laboratory (and
less expensive) computer.
To begin the «imiitotrôn, we assumed a symmetrical
structure (same alignment for the two rings), with the
principal axis OZ roughly along the elongation axis. Two
chemical shifts were assumed, namely one for the phenyl
protons and one for the methyl protons. The starting
value of the relative chemical shift &o was fixed to that
measured (with a resolution of ~40 Hz) in the isotropic
priase, namely AOJ» == 4.9 ppm, that is, 440 Hz for spectra
taken at 90.14 MHz. The result of the simulation was
very bad, whatever the dihedral angle between the rings.
What could never be reproduced was the broad central
pan of the spectrum. It turned out that this situation
occurred because the assumed structure was too sym-
metrical. Thefurtherstepwasthustodistortthemolecule
so that the alignments of the two rings were different. The
situation improved significantly, but now the problem was
to know how to describe exactly the distortion.
Wethus turned toasimpler problem, namely what could
\ be learned about the relative orientations of the two rings
,3K.
Macromolecules. Vol. 25, No. 18,1992
(Ql
(b)
Mlà
(C)
Figure 3. Simulated proton NMR spectra of the six-spin system
constituted by ring A (a) and by ring B (b); average of spectra
a and b (c); exact simulated spectrum of the twelve-spin system
(d). Theparametersarethoeeforthebestsimulationofspectrum
Ic.
by ignoring the interactions between the rings, that is, by
treating the problem as a two six-spin system constituted
bythetwomethyl-substitutedrings. Thisapproachtumed
out to be very fruitful as it showed that the main features
of the central part of the spectrum were mainly due to the
intra-ring interactions. The interactions which played a
major role in this analysis were the relative values of the
ortho interactions D» and 07$ and that of intramethyl
interactions Du and #10.11' a prerequisite for a good
simulation was that these two hitter interactions differ by
roughly a factor of 2. After many attempts, it turned out
that the least aligned ring was ring B (as in PAA15!) and
that the corresponding methyl group was associated with
the largest intramethyl interaction in absolute value. These
findings gave hints on how the distortions, described by
the various Euler angles mentioned above, should be
chosen. These results are illustrated in Figure 3a-c, where
are shown the two six-spin spectra corresponding to the
two rings, calculated using values of the interactions of
ths final solution given below, as well as the average of
tluoe two spectra. Comparison of Figure 3c with the
experimental spectra immediately reveals that the main
features in the center of the spectrum are indeed due to
interactions inside the six-spin systems. In particular,
the doublet of sharp Unes closest to the center of the
spectra, well visible in the more symmetrical spectra of
the polymers Ia and Ib, is essentially due to the intra-
methyl interaction DIO.U of the more aligned ring A.
However, comparison with Figure 1 shows that the two
times six spin approximation is a bad approximation to
reproduce satisfactorily the experimental spectra.
The last part of the study thus consisted in introducing
the inter-ring interactions and using the simulated (and
expensive) twelve-spinspectrum to refine the simulations.
The procedure was as follows: the values of the ten most
important interactions were guessed to give the best
simulations. Thesevalueswerethenintroducedinafitting
program to find the values of the geometrical (and
motional) parameters which allow these interactions to

143
Moeromolecules, VoL 25, No. 18,1992
be reproduced. This geometry was then used to calculate
all other interactions, and the spectrum was simulated
again with these final values. The chemical shift was
introduced as an asymmetry parameter of the spectrum,
defined as the ratio of the (unknown) value of the shift to
the main interaction DTS- This parameter cannot be
considered on the same level as the other ones since it is
determined by the position of the central pan of the
spectrum with respect to the large external doublet This
parameter is easily converted into the value of a chemical
shift Air (in ppm) by taking into account the actual width
of the particular experimental spectrum considered.
This fitting procedure was a rather lengthy trial and
error work. However, since the starting point was suffi-
ciently close to the final solutions, convergence could be
reached. The result is shown in Figures 3d and lc,d for
AcSAcdS and in Figure Ia for polymer DDA9d20, which
correspond to two extreme situations. In Figure Ib is
shown the theoretical spectrum corresponding to the
geometry of polymer DDA9d20, but with a slightly larger
chemicalshift ThespectrumcfpolymerAZA9dl4clearly
appears to be intermediate, although closer to that of
DDASdZO. Thus two structures can he proposed, one for
AcSAc and one for the polymers. It is not very useful to
describe the various attempts that have been madetoreach
these solutions, but it may be worth mentioning the
following:
(i) The two Euler angles describing the orientation of
each methyl group within its ring frame played a com-
plementary role—namely combinations of the two pa-
rameters gave identical results. Since the average value
of the inclination crossed the value 60° (or 120°), we
somewhat arbitrarily fixed these inclinations to these
values and determined the orientation angles vfe for the
two groups es those yielding the best simulation. lathis
way, the number of parameters is reduced to ten as
mentioned above.
(ii) The best fits to the exp^rimentalspectraareobtained
with different values of the relative chemical shift This
shift decreases as the absolute width of the spectrum
(proportional to So) increases (see below).
(iii) The fact that the fits to the experimental spectra
of Ac9Acd6 (Figure led), for which there are a priori no
broadening effects due to polydispersity or slow motions,
are not perfect may have several causes, including (a)
approximations made in the modeling, such as the choice
of two different chemical shifts instead of at least eight,
as suggested by the high-resolution spectra Ui the isotropic
phase, the assumption that the azoxy moiety CNMC is a
perfect plane, or even limitations of the single-conforma-
tion model itself, (b) inherent uncertainties associated with
the fast Fourier transformation of an experimental free
induction decay signal to obtain perfect line shapes,
whatever the quality of the NMR spectrometer and the
care with which the experiments are made (very broad
spectral range, subtraction of an empty tube signal, etc.),
find (c) imperfection of the partial deuteration, implying
the contribution to the signal of nondeuterated and/or
differently partially deuterated molecules. In view of all
this, we can consider that the fits are satisfactory.
(iv) Finally, it is worth mentioning that introduction of
biaiklity in Ae modeling does not improve significantly
the simulations, in theaense that the effect of introducing
some finite value of lSXx ~ SyyVSzz can be very easily
compensated by slightly changing the geometrical pa-
rameters. This means that no significative information
about the biaxiality can be extracted from these simula-
tions only, except that it is probably weak, in agreement
- .«*•:
NMR Spectrum of a Twelve Spin V: System 4523
with the results of ref 2.
In the Appendix are given the values of the 30 dipolar
interactions corresponding to the best simulated spectra
of Ac9Acd6 and of DDA9d20 (shown in Figures Ic, d, and
Ia, respectively).
5. Results
The optimized geometrical and motional parameters
corresponding to the simulated spectra of Ac9Acd6 and
of DDA9d20 polymer (in parentheses for the latter) are
given below with all their numerical figures to allow the
interested reader to repeat the calculations. The sys-
tematic errors are estimated to —0.1° for polar angles and
~1° for azimuthal angles. For AZA9dl4 polymer, the
values are intermediate, although closer to those for
DDA9d20 polymer: (i) Orientation of frame B with respect
to the principal frame of the order tensor:
U1, irrelevant due to uniaxiality
V1 = 12.107° (11.242°)
W1 = 82.238° (70.660°)
(U) Orientation of frame /3 with respect to frame B:
U2 = -54.257° (-63.939°)
V.j =-10.210° (-11.576°)
W2 =15.642° (26.706°)
(iii) Orientation of frame A with respect to frame a:
U3 = -33.180° (-32.695°)
V3 = -2.595° (-2.491°)
W3 = 7.559° (12.550°)
(iv) Orientation of methyl groups around the threefold
axis, assumsd to be inclined at 60" (120°)on the para axis:
„/ = -5.060° (-3.587°)
Vn? = -26.69° (-1.272°)
From the above numbers, any geometrical quantity can
be calculated. For example, the angle between the two
para axes is 10.014° (11.568°). The value is 13.5° in solid
PAA.15 Thisanglebeingsmau,thedihedralanglebetween
the two rings is given approximatively by D = |I/2 + W2
+ U3 +W^ = 64.24° (57.38°) to be compared with 64.32°
(57.82°) for the angle15between the two normals to the rings
(22.6° in solid PAA -20). The two dihedral angles with
the intermediate azoxy plane are \U2 + W2] = 38.61°
(37.23°)forringBand|{/3+ W3| = 25.62° (20.14°) for ring
A (19.7 and 3.0°, respectively, in solid PAA20).
It is interesting to note that these results are fully
consistent with theoretical calculations on the isolated
molecule22made with the quantum semiempirical method
PCILO: the values found for these dihedral angles
correspond to a point which is inside (or, for the polymer,
very close to) a zone where the intramolecular energy is
the lowest, namely zone I in the energy map represented
in Fife are 2 of ref 22-

Figure 4. Stereoscopic projections of the mesogenic moiety of
Ac9Acd6 dissolved in nematic PAAdU in two perpendicular
planes containing the principal axis OZ, assumed to be vertical.
These two planes correspond approximately to the two bisector
planes of the phenyl rings.
Same «s Figure 4 for DDA9d20 polymer. Note that
the moiety is globally more aligned along OZ than in AcSAcdS.
Itis observed thatthe différences between the structure
of AtSAe (the monomer) in solution in PAA and the
structures of the polymers in their neat nematic phase are
rather subtle and lie in slight changes of the distortions,
of less than 2° for polar angles and of a few degrees for
azimuthal angles (except for the a»*). Although small,
v! these differences are sufficient to produce measu .jle
changes in the spectra (because several angles are close to
the magic angle).
Figures 4 and 5 show stereoscopic projections of the
moiety in two perpendicular planes containing the prin-
cipal axis OZ, and which roughly correspond to the two
bisector planes of the phenyl rings (close to the two
principal planes in a situation where the order is biaxial),
forAc9AcandiorDDA9d20. It is observed that the me-
sogenic moiety a more aligned along the principal axis in
the polymer than in the monomer, aad this result may be
144
Macromoleeules, Vol. 25, No. 18,1992
regarded as the contribution of the spacers (i.e., of the
polymeric nature of the molecule) to the stretching of the
mesogens.
The proportionality constant between the main splitting
of the spectrum 2iN as defined in Figure 3d and the main
order parameter Szz (=So in the case of uniaxiality) is
easily determined from thesimulationbysimplymeasuring
the width of the simulated spectra. We obtain for the
polymers
26N/kHz = (22.52 =F 0.30)S2Z (1)
The error bar includes an uncertainty of ±0.01 A on the
distance between the ortho protons (the value of 2.475 A
b chcaen in the simulations). It is worth noting that the
proportionality constant of 22.74, assumed in previous
works2 and deduced from an approximate simulation of
the spectrum of the aondeuterated DDA9 polymer;21 is
consistent with the present results and can thus still be
used to estimate order parameters hi spacer-deuterated
ME9Sn polymers (for normal polymers, the constant is
larger, =24.06 (ref 2)). For Ac9Acd6 in PAAdM, the
simulation yields a slightly smaller central value of 22.31.
Finally, the results concerning the chemical shifts ACT
may be summarized as follows. For Ac9Acd6, the un-
certainty on Air deduced from the simulations is =±0.3,
whereas for polymer DDA9d20, it is much larger, »1.5. It
is easily shown that for iininiinl order Air is given by (see,
o.g., eqs 1-4 of ref 4)
Aa = A<rto - (Affto - AS2Z)S0 (2)
where AS^z is the component, along the principal axis
OZ, of the difference between the chemical shift tensors
of the phenyl and methyl protons. With AiTjx, = 4.9 ± 0.3
ppm, a least-squares fit of eq 2 to the three points for
AcSAcdfi yields ASzz=1.8 ± 0.9 ppm. Although we learn
here that this anisotropic component is positive and
smaller than a?*,, this value by itself corresponds to very
little information on the anisotropic part of this tensor.
6. Comparison with PAA
Besides the weak differences between the results for
AcSAc (monomer) and Ae polymers already mentioned,
the other interesting result of this study is the similarity
with PAA, despite the presence of the two methyl groups.
The similarities lie not only in ths structure but also in
the order.
For the structure, the most important similarity is the
distorted character of the backbone constituted by the
azoxy group and the two para axes of the rings. Apart
from the absence of r-flips of the rings (which, incidently,
have allowed such accurate information on the structure
to be obtained due to nonaveraging of important inter-
actions), the only significant differences lie in the larger
values of the dihedral angles between the planes of the
three moieties. However, as in PAA, the angle of the azoxy
plane with ting B is larger than with ring A.
For the order, the similarities with PAA are even larger.
Ring B is the least aligned ring. The inclination of its
para axis on the principal axis OZ is [CB! = Vi = 12.107°
(11.242° for the polymer). The inclination of ring A is
easily calculated, and its value is M = 4.168° (»0.3° for
the polymer). In nematic PAA, reasonable values are |<B|
» 12° and foj=4.5".1"5 This shows that the orientations
of the principal axis witLi^pect to this backbone are very
similar in all molecules (practically the same for AcSAc
and PAA).
 145

Macromolecules, VoI 25, No. 18,1992 NMR Spectrum of a Twelve Spin Vz System 4525

All these similarities aie probably to be associated with the (magnetically equivalent) spins of the methyl groups
the high stability of the chemical structure of the atoxy are labeled m for B and n for A.
group.
7. Conclusion intrunoiety B intztmoiety'A intennoietiea B-A

This study constitutes a further example which shows ij n» •V Da ij Ai

that simulation of (one quantum) proton NMR spectra in mja -1.8569 (-2.2802) n,n -«3649 (-1.3777) m.7 0.0451 (0.0430)
anisotropic media can be a powerful method to determine n>,4 -3.1715 (-3.0499) 9Ji -3.6075(-3.58M) m.8 -0.0693 (-0.0682)
m.5 0.1018(0.1250) 8ji 0.0529 (0.0755) m.9 -0.2357 (-0.3042)
the structure of rather complex nonrigid molecules in a B.6 0.4388(0.4444) 7,D 0.4180 (0.4329) nwi -0.3184 (-0.7475)
liquid phase. This has been possible because a simple 4.5 0.7606 (0.7529) 8,9 0.7608(0.7623) 4,7 -0.1308 (-0.1287)
theoretical framework has been chosen (the single- 4.6 0.1036 (O.OS18) Iff 0.1591 (0.1178) 43 -0.0867 (-0.0869)
conformation model) and because exact simulation has 5.6 -7.4002 (-7.4713) IJt -7 .8602 (-7.9225) 4,9 -0.1276 (-0.1439)
4,u -0.2715 (-0.3507)
been performed. To our knowledge, this is the first exact 5.7 -0.4047 (-0.4084)
simulation with such a large number of spins (12). The 5,8 -0.1697 (-0.1685)
structures, conformations, and orders of the aromatic core 5,9 -0.1419 (-0.1372)
of Ac9Ac and of ME9Sn polymers could be determined Sf -0.2505 (-0.2245)
6,7 -0.8088 (-0.7869)
with reasonable accuracy, and analogies and differences 6,8 -0.3550 (-0.3438)
with PAA could be put in evidence. 6,9 -0.3391 (-0.3163)
The most important result for the molecular physics of 6,n -0.3523 (-0.2732)
(polymer) liquid crystals that emerges from this study is
Ae fact that the changes in the structure and order from References and Notes
monomer to polymer are rather weak. In particular, it (1) Volino. F.; Ratio, J. A.; Galland, D.; Esnault, P.; Dianouz, A.
appears that, in these nematic polymers at least, the ori- J. Mot. Cryst. Lu1. Cryst. 1990,191,123.
entational characteristics (essentially the orientation of (Z) Esnault, P.; Galland, D.; Volino, F.; Blumstein, R. B. Macro-
the principal axis of the mesogen) are mainly determined moleeules 1989.22, 3734.
by the nature of the mesogen and little by the length of (3) Blumstein, A,; Thomas. O. Macromolecules 1982,15,1264.
the aliphatic spacer or the polymeric nature of the (4) Blumstein, A.; VUasagar, S/, Ponrathnam, S.; Clough, S. B.;
Maret, G.; Blumstein, R. B. J. Polym. Sd., Polym. Phys. Ed.
molecule. Thia result will be an important ingredient for 1982,20,887.
the analysis of the deuterium NMR data of the corre- (5) Blumstein, R. B.; Blumstein, A. Mat. Cryst. Liq. Ciyst. 1988,
sponding spacer-deuteraisd dimers and polymers. Cor- 165,361.
responding preliminary results have been presented in ref (6) Esnault, P^ Casquilho, J. P.; Volino, F. Liq. Cryst. 1988, 3,
1. 1425.
(7) Esnault, P.; Gauthier, M. M.; Volino, F.; d'Allest, J. F.; Blum-
stein,R.B.Mol.Co»t.£io.Cryst. 1988,157,273. See also the
Acknowledgment. We are indebted to Professors R. two preceding papexs in the aame issue.
B. and A. Blumstein and to Dr. J. F. d'Allest for providing (8) Blumstein, R B.; Stickles, E. M; Gauthier, M. M; Blumstein,
the polymer samples as well as the basic molecule for A^ Volino. F. Hacromolecules bit, 17,177.
synthesis of Ac9Acd6. We also thank Dr. A. M. Giroud- (9) Esnault, P.; Volino, F.; Gauthier, M. M.; Giroud-Godquin, A.
M. MoL Cryst. Liq. Cryst. 1986,139,217.
GodQuinforsynthesisofPAAdU. Thisworkwaspartially (10) Cochin, M.; Pines, A-; Rose, M. E.; Rucker, S. P.; Schmidt, C.
supported by NSF/CNRS Grant INT-871501/88-920156 WoL Phys. 1990,69,671.
and NSF Grant DMR8600029. (11) Limmer. St.; SehmiedeL 64 HiIInM, B.; Losche, A-; Grande, S.
J. Phys. Fr. 1980,41,869.
Appendix (12) Komolkin, A- V.; Molchanov, Yu. V. (a) IAq. Cryst. 1989,4,117;
(b) MoI Cryst. IAq. Cryst. 1990,192,173.
Here, we give the values (in kHz) of the 30 dipole-dipole (13) Ferreira, J.B.;Martin^A-F.;Galland,D.;Volino,F.Afo!.CrysI.
interactions Dg which correspond to Ac9Acd6 in PAAdU Liq- Cryst. 1987, JSI, 283.
andtoDDA9d20polymer. These interactions are defined (14) Galland. D.; Volino, F. J. Phys. Fr. 1989,50,1743.
by* (15) Volino, F.; Galland, D.; Ferreira, J. B.; Dianoux, A. J. J. MoL
Liquids 19S9,43, 215.
(16) Galland, D.; Volino, F. J. Phys. II1991,1,209.
(17) Volino, F. Proceedings of the European Conference on Liquid
Crystals, Courmayeur, Italy 1991. MoI. Cryst. Liq. Cryst., in
r v m press.
(18) Ciajolo, M. R.; LeIj. F^ Tancredi, T.; Temusai, P. A.; Tim. A.
where rç is the distance between protons i and j and fly- Acta Crystallogr. 1X1,838, ra2B.
is the angle between the vector ij and the static magnetic (19) Kaftory, M; Rubin, M. B. J. Chem. Sx., Perkin Trans. 21983,
field. They correspond to the particular case where the 149.
orientations! order is uniaxial and perfect (S0 = 1). The (20) Krigbaum, W. R.; Chatani, Y.; Barber, P. G. Acta Crystallogr.
numerical value ofthe constanth-rp'M*2 is 120.12 kHz-A3. 1970,B26,97.
(21) Martins. A. F.; Ferreira, J. B.; Volino, F.; Blumstein, A.; Blum-
The numbers correspond to Ac9Acd6. For polymers stein, R B. Uacromolecules 1983,16, 279.
DDA9d20, they are given in parentheses. For simplicity, (22) Berges, J.; Perrin, H. MoI. Crysl. Liq. Cryst. 1984.113, 269.
 CHAPITRE III

POLYMERES

CONVENTIONNELS

164
/is
 149

ni A Simulation de spectres RHN d'un polymère rigide

m A 1 Généralités

Contrairement aux travaux précédents qui ont déjà donné lieu à

une publication, nous allons développer dans ce troisième chapitre
l'application des deux modèles (conformation unique et influence des
modes élastiques) à deux polymères, l'un rigide et l'autre
semi-flexible, ne présentant pas de phase nématique lorsqu'ils sont
purs.
Le (poly[oxy(chloro-l,4-phénylène)oxycarbonyl] [trifluorométhyl)
-1,4-phénylène] carbonyl) (PTFC), est un polyester présentant une
transition vitreuse nématique à 71°C et une phase nématique en
solution dans le PAA[25]. Sa formule est la suivante:

avec n = 21 pour les échantillons utilisés dans cette étude.

Le but de celle-ci est d'appliquer le modèle à conformation
unique afin de reproduire les spectres de RMN du proton et du fluor en
phase nématique du mélange PTFC/PAAdl4, puis d'en vérifier les
résultats sur la phase pure du polymère.
Le benzoate de phényl (molécule parente du PTFC dont aucun des
deux groupes phényl n'est substitué) a fourni les données de base[26] '
quant à la structure géométrique de l'unité monomère de PTFC qui
constitue le système à 9 spin 1/2 (six protons et trois fluors),
supposé (suffisamment) isolé, dont on a calculé les interactions ,
magnétiques. Les valeurs des divers angles et distances utilisés pour
les simulations sont indiquées sur la figure 5. Les deux groupements
phényls sont supposés être des hexagones parfaits, reliés entre eux
par un groupement supposé plan. Le plan de ce groupement est
f* désor.enté de 6.2" par rapport au plan du phényl fluoré, le plan du 5t
I phényl chloré étant lui-même incliné de -68.4° par rapport à 1
t ; celui-là[26]. D'un point de vue géométrique, seule l'orientation du \•[
; groupement trifluorométhylène vis-à-vis du phényl, ainsi que sa "i
structure, ne proviennent pas des données du benzoate de phényl. Cette
orientation a été estimée à 121° (ou 61° selon la position du CF3) par
( rapport à Taxe para du phényl, les autres données géométriques ayant t
» été fixées à des valeurs conventionnelles[27]. En ce qui concerne A
J? l'ordre or i entât ionnel, l'axe principal Z du tenseur d'ordre a été <f
1 supposé, conformément aux résultats obtenus dans le cas du PAA[Sa], »,
Parallèle à la liaison C-O attachée au phényl chloré. , j
 150

figure 5 : Unité monomère du PTFC. Les angles et distances indiqués

sont ceux utilisés pour le calcul des spectres RMN
théoriques du proton et du fluor.

Les seules inconnues considérées dans le problème sont donc

l'emplacement, sur les .phényls substitués, du chlore et du groupe
trifluorométhyl, ainsi que l'existence possible de mouvements de
rotation de IT de ces phényls substitués autour de leur axe para.

1 A 2 Spectres alignés du polymère en solution dans un solvant

nématique

Le calcul exact, grâce aux données et hypothèses évoquées plus

haut quant à la structure géométrique et à l'ordre associé au PTFC,
montre que le spectre théorique de cette molécule se décompose en deux
blocs. L'un est centré autour de la fréquence de Larmor du fluor (on
l'appellera, par la suite, spectre fluor) et l'autre autour de la
fréquence de Larmor du proton (il sera appelé spectre proton).
Les simulations effectuées et leur comparaison avec les spectres
du fluor expérimentaux en phase nématique permettent d'écarter
l'hypothèse d'une rotation rapide du phényl fluoré, comme on peut le
constater sur la figure 6, mais ne permettent de conclure ni sur la
position du chlore et du groupe CF3, ni sur la rotation éventuelle du
phényl chloré.
Les spectres de RMN du proton apportent beaucoup moins
v d'information. Tous les spectres obtenus, quels que soient les
hypothèses envisagées pour la rotation du phényl chloré ou la position
\ du chlore, sont, sinon équivalents, du moins aussi proches des
 151

spectres expérimentaux. La concordance avec ceux-ci est d'ailleurs

bien moindre que dans le cas du fluor, comme on peut le voir sur la
figure 7. La comparaison semble cependant légèrement favoriser la
position du groupe CF3 en ortho plutôt qu'en meta. Enfin, les deux
séries de spectres expérimentaux ne permettent de déterminer si
Tordre associé au mélange nématique est biaxial ou non.
Tous les spectres expérimentaux ont été enregistrés sur un
spectromètre BRUCKER CXP90, avec des fréquences de Larmor de 90.1 et
84.8 MHz pour le 1 H et le 19F respectivement

b)

10 kHz

c)

5
I
Vj

figure 6 : Spectres fluor du PTFC a) spectre théorique calculé en

supposant une rotation rapide de TT du phényl
trifluorométhylé b) spectre théorique calculé sans
rotation du phényl trifluorométhylé c) spectre
expérimental du PTFC en solution dans le PAAdl4 (24%
massique) à 133° C. Les spectres théoriques ont une
largeur totale de 40 kHz.
 152

a)

figure 7 : Spectres proton du PTFC a) spectre théorique calculé en

supposant une rotation rapide de ir du phényl chloré b)
spectre théorique calculé sans rotation rapide de TT du
phényl chloré c) spectre expérimental à 1330C du PTFC en
solution dans le PAAdl4 de concentration massique 24% et
d) 5%. Les deux spectres théoriques ont une largeur totale
de 40 kHz et correspondent au cas où le phényl fluoré
n'effectue pas de rotation rapide autour de son axe para.
 153

Ces différences flagrantes entre les spectres proton calculés et

expérimentaux, alors qu'elles sont difficilement perceptibles dans le
cas du fluor, peuvent avoir plusieurs causes:
a) la polydispersité du polymère ainsi que 1'inhomogénéité du mélange
peuvent être responsables d'une distribution de concentrations et de
paramètres d'ordre au sein de l'échantillon. Cette supposition est
supportée par l'évolution de la forme de raie en RMN du proton en
fonction de la concentration en PTFC, le spectre expérimental se
rapprochant des divers spectres théoriques à mesure que cette
concentration augmente ainsi qu'on peut le remarquer sur la figure 7.
Cette distribution se remarquerait moins sur les spectres fluor en
raison de l'anisotropie de déplacement chimique relativement plus
importante (on peut néammoins, sur le spectre expérimental de la
figure 6, distinguer un léger renflement, à gauche du centre du
spectre, pouvant résulter de cette distribution).
b) le fait de ne considérer que l'unité monomère dans le calcul des
interactions peut être trop approximatif. En effet, les spin 1/2 des
monomères voisins peuvent intervenir de façon non négligeable sur le
spectre proton, tout en ayant peu d'influence sur le spectre du fluor,
le groupe CF3 ayant en position ortho une position assez centrale au
sein du monomère. D'autre part, tous les paramétras géométriques ont
été fixés dans cette étude, la plupart d'après les données disponibles
sur le benzoate de phényl. La conformation de l'unité monomère du PTFC
peut bien sûr être, et est certainement, légèrement différente de
celle supposée, mais de légères modifications des angles et distances
considérées ne permettent pas d'observer d'amélioration suffisamment
marquante pour être retenue. Il en est de même pour l'orientation de
Taxe Z du tenseur d'ordre orientationnel dont une variation de
quelques degrés influe assez peu sur les formes de raies observées.
Néammoins, ces résultats obtenus avec pour seuls paramètres les
mouvements internes et la position des substituants sont
(surprenament) bons et permettent de considérer que le modèle à
conformation unique s'applique très bien au cas du PTFC en mélange
dans le PAAdl4.

m A 3 Spectres du polymère pur

Un moyen d'éclairer ces résultats sous un angle différent est de

les utiliser afin de simuler les spectres du proton et du fluor du
PTFC pur disponibles entre 115 et 1950C. Le principe de c-^tte
simulation est le suivant on considère que le PTFC pur, à ces
températures macroscopiquement amorphe, est composé d'une multitude de
monodomaines nématiques dont les vecteurs directeurs sont
isotropiquement dist-ibués {la taille de ces monodomaines pouvant
aller de quelques inolécules à des groupes plus importants mais de
taille suffisamment réduite afin de pouvoir négliger l'influence
 154

éventuelle de modes visco-élastiques lents). Le spectre ré it

correspond à la moyenne sur les angle? Q (0 étant l'angle entre le
directeur du monodomaine et le champ statique) des spectres théoriques
(en travaillant dans l'hypothèse uniaxialê, le spectre associé à ce
monodomaine est alors le spectre théorique, rétréci d'.un^facteur S qui
traduit le désordre vis-à-vis du directeur et d'un facteur P2(cbs®)
traduisant la désorientation du directeur par rapport à l'axe de
mesure, et centré en V0 + 2/3 Aa P2(cose), où V0 est la fréquence de
Larmor et ûo- = CTU - ^1 l'anisotropie de déplacement chimique)[16,27],
chaque angle @ étant équiprobable. Les spectre" théoriques considérés
pour chacun de ces monodomaines sont ceux correspondant aux hypothèses
suivantes: absence de rotation des deux phényls autour de leur axe
para, chlore en position meta et CF3 en position ortho, ordre
uni axial.
L'anisotropie de déplacement chimique ACT, négligeable dans le cas
du proton, a du être déterminée pour le fluor. Ceci a été permis grâce
aux spectres fluor de la biphase N+I du mélange PTFC/PAAdl4. La partie
alignée des spectres proton et fluor nous a permis de déduire, par
comparaison avec les spectres théoriques calculés dans l'hypothèse
d'un ordre uni axial, le paramètre d'ordre S de cette phase nématique,
qui vaut 0.73. Ce spectre étant translaté en fréquence d'une valeur
Sx2/3xAa par rapport au pic de la phase isotrope (S=O), il est alors
facile de déduire, d'après la figure 8, que 2/3 âa = -479 ± 5 Hz.
Avec un paramètre d'ordre uniaxial S égal à .8, on reproduit de
manière quasi-parfaite les spectres expérimentaux du proton et du
fluor à haute température (entre 115 et 195°C), au moyen néanmoins
d'un élargissement du spectre par une convolution par une lorentzienne
de demi-largeur & à mi-hauteur. Les simulations obtenues peuvent être
comparées aux spectres expérimentaux sur les figures 9 et 10, les
valeurs correspondantes de & utilisées étant données sur la figure 11.

V
1 figure 8 : Spectre fluor du PTFC dans la biphasé N+1 de son mélange
avec le PAA (24% massique) à 161°C. La partie alignée du
spectre correspond à un S uniaxial de l'ordre de 0.73.
 155

6 = 1400 Hz

T = 1160C
6 = 3500 Hz

figure 9 : Spectres fluor expérimentaux (à gauche) et simulés du PTFC

pur. Les spectres expérimentaux ont une largeur totale de
44.3 kHz.

S - 900 Hz

S = 1800 Hz

Si*

Vi

figure 10 : Spectres proton expérimentaux (à gauche) et simulés du

,1? PTFC pur. Les spectres expérimentaux ont une largeur
totale de 42 kHz.
 156

200 127 100 T(0C)

Ln(6/Hz)i 6/kHz
-10
9-1

-5

8-
-2
,4+
7-
-1
1
H
-.5
i i i I I i * i •
2 2.5 33 - 3
10 /T(K'1)
figure 11 : Variation du paramètre d'élargissement 6 du PTFC pur en
fonction de la température.

Ce facteur d'élargissement 6, croissant lorsque la température

diminue peut s'interpréter comme l'effet des mouvements de translation
qui. jusqu'ici, ont été considérés comme suffisamment rapides dans les
pL- = :.es nématiques pour annuler les interactions intermoléculaires.
Ce-:: n'est peut-être plus le cas dans le cas du PTFC pur, les
mouvements de translation ralentissant quand la température décroît ce
qui explique l'augmentation de la valeur de 6.
Pour ces simulations de spectres du PTFC pur, on constate une
tendance inverse à celle observée en phase nématique. Toutes les
simulations reproduisent exactement les spectres expérimentaux sauf
î dans le cas des spectres du fluor au delà de 140°C. Ces spectres
présentent en effet un pic central bien plus prononcé et bien mieux
résolu que celui obtenu par les simulations en utilisant un S
permettant de reproduire parfaitement les spectres proton à des
températures équivalentes. D'autre part le paramètre S ne semble pas,
ou très peu. varier sur la gamme de température étudiée, restant à une
valeur proche de .8.
 157

Cette différence entre spectres expérimentaux et calculés peut

encore provenir de l'incertitude sur le spectre de base du monodomaine
que l'on considère. En effet, le couple de spectres choisi ( 1 H 5 19 F)
donne, pour les spectres proton, un résultat parfait à toute
température pour le PTFC pur mais c'est sur ce même spectre proton que
l'incertitude était la plus grande dans l'étude du mélange
PTFC/PAAdl4, la forme de raie théorique en fluor étant assez
précisément déterminée. C'est donc, a priori, aux résultats obtenus
par l'étude des spectres 19 F du PTFC que l'on devrait se fier, la
différence observée à haute température provenant soit du phénomène de
polydispersité précédemment évoqué, soit de l'influence des modes
élastiques qui, si elle pouvait être négligée dans le cas du PTFC en
solution dans un nématique de faible masse moléculaire, peut s'avérer
cruciale dans le PTFC pur. Ces deux phénomènes affectant également le
spectre proton, il devient implicite que le spectre de base considéré
en proton n'est pas adéquat.

Malgré ces imperfections, l'étude des spectres du PTFC dans la

phase nématique de son mélange avec le PAA et dans sa phase amorphe a
abouti à plusieurs résultats dont certains ne laissent place à aucune
incertitude. Parmi ceux-ci, en premier lieu, l'absence de rotation
rapide de ir du phényl fluoré autour de son axe para est clairement
établie par la comparaison des spectres de RMN du fluor théoriques et
expérimentaux en solution dans le PAAdl4. Malheureusement, il n'est
pas possible de discerner si une telle rotation existe ou pas pour le
phényl chloré, ni de prédire la position des substituants, ces
paramètres ayant une influence moindre, quoique notable, sur les
spectres calculés qui ne permet pas de trancher par comparaison aux
spectres expérimentaux soumis à des effets telle que la
polydispersité. Cependant, et compte tenu du faible nombre de
paramètres laissés libres, les résultats convaincants obtenus tant
pour la simulation des spectres de la phase nématique du mélange
PTFC/PAAdl4 que pour celle de la phase amorphe du PTFC pur, plaident
en faveur du modèle utilisé.
 159

i B Simulation de spectres RHN d'un polymère semi-flexible

Les résultats obtenus pour le PTFC, incitant à poursuivre

l'interprétation des spectres RMN de polymères non nématiques en tant
que spectres de poudre de monodomaines nématiques, ont conduit à
analyser les spectres de RHN du 19 F du polytétrafluoroéthylène (PTFE)
de composition (CF2Jn , enregistrés entre 360 et 460 K, selon des
principes identiques.
Les spectres théoriques du PTFE parfaitement aligné ont été
calculés à partir des données cristallographiques fournies dans [29],
pour des bouts de chaîne comprenant huit à douze 19F. La cc:iformation
utilisée dans ces calculs est celle du PTFE cristallin à haute
température (entre 19 et 3O0C). Les atomes de carbone sont alors
disposés sur une hélice IB7 avec une distance de répétition de 19.5 A
(sur cette distance, l'hélice fait sept tours complets et comporte
quinze atomes de carbone). La conformation considérée est donc la
suivante: chaque groupement CF2 est contenue dans le plan bissecteur
des deux liaisons C-C auxquelles appartient le carbona, chaque plan
CCC faisant un angle de 168" avec le précédent au fur et à mesure que
l'on avance selon l'axe z de l'hélice. Ceci, ainsi que les distances
et angles utilisés[30], est décrit dans la figure 12. D'un point de
vue orientât!onnel, il a été supposé que l'axe principal Z du tenseur
d'ordre est selon l'axe de l'hélice et que Tordre est uniaxial. Le
spectre résultant de ces calculs varie sensiblement en fonction du
nombre de groupements CF2 considérés, comme on peut le voir sur la
figure 13. La prise en compte des mouvements de racémisation
(l'échange entre les deux 19 F d'un groupement CF2) semble par contre
avoir de moins en moins d'influence à mesure que l'on augmente ce
nombre.

'I

figure 12 : Structure du groupement CF2, et de son entourage, utilisée

dans les simulations.
 160

a)

^iflM^

figure 13 : Spectres calculés du PTFE avec respectivement a) quatre b)

cinq et c) six groupements CF 2 . Ces spectres tiennent
compte de mouvements rapides de racémisation. Leur largeur
totale est de 44.4 kHz.

Afin de calculer des spectres de poudre, il est également

nécessaire de connaître Tanisotropie de déplacement chimique Au
associée aux groupements CF2 du PTFE. Ce ACT intervient comme décrit
précédemment dans le cas du PTFC dans le calcul. Le Aa utilisé
provient des données de Garroway et al. [31] et a une valeur de
8.83616 khz pour nos expériences. Une autre valeur, de 9.0736 kHz, est
fournie par Vega et al. [32].
i La méthode utilisée pour le PTFC, c'est-à-dire le calcul de
spectres de poudre de monodomaines nématiques caractérisés par un
paramètre d'ordre uni axial S et éventuellement convoiués par une
lorentzienne afin de tenir compte des interactions intermoléculaires,
ne permet pas de simuler correctement les spectres du PTFE pur entre
l'ambiante et 20O 0 C. Néammoins, les résultats obtenus aux plus basses
températures semblent favoriser le ACT de Garroway et al. [31]. Ceci
V peut être la conséquence de plusieurs causes: i) le spectre théorique
1 utilisé ne comprend pas suffisamment de groupes CF2 (6 ont été pris en
compte dans notre cas) et ne correspond donc pas au spectre d'un
 161

roonodomaine de PTFE aligné ii) les modes élastiques, ignorés dans le

cas du PTFC, ont ici une influence notable sur la forme de raie ou
alors iii) le modèle n'est pas adéquat pour le PTFE. Les résultats
encourageants, quoiqu'imparfaits, obtenus pour le PTFC nous ont
incités à explorer la deuxième hypothèse.
La théorie exposée au premier chapitre a donc été appliquée, avec
une distribution F(0,<t>) = l/4ir et le spectre de base calculé pour
douze 19F, en y supposant des mouvements de racémisation assez rapides
et un ordre uni axial, le spectre résultant étant ensuite convoiué,
toujours afin ce tenir compte des interactions intermoléculaires. Ceci
permet de reproouire les spectres expérimentaux du PTFE, comme on peut
le constater sur la figure 14, depuis des température proches de
l'ambiante (~ 15°C) jusqu'à ~ ZOO0C, pour des valeurs de S et TC, dans
le cadre de l'approximation sphérique à une constante, allant
respectivement de .75 à .2 et de 1. à 10~5s.

a)

figure 14 Spectres fluor expérimentaux (à gauche) et simulés du PTFE

pur à a) 187°C et b) 17°C. La largeur totale du spectre
expérimental est 40.7 kHz pour le spectre à haute
température et 82.6 kHz pour le spectre à température
ambiante.
 162

Cette décroissance du S quand la température augmente n'est

nullement étonnante, même si l'on arrive, au plus hautes températures,
à une valeur très faible de .2, cette valeur s'accordant assez mal
avec l'hypothèse de petites fluctuations sous-tendant le modèle. En
supposant des mouvements internes à la molécule de base, associés à
une biaxialité éventuelle de Tordre dans les monodomaines nématiques
de PTFE, on peut alors effectuer les calculs à partir d'un spectre de
base réduit d'un facteur P2(cosx) (le paramètre d'ordre interne à la
molécule de base) et obtenir des simulations avec un S uniaxial plus
élevé, mais restant inférieur à .4, le rc augmentant conformément à la
loi d'échelle observée au chapitre I et la qualité de la simulation
diminuant à mesure que S augmente. La franche augmentation du TC quand
la température diminue implique, selon (21), une forte augmentation de
la valeur de -n/K à mesure que Ton se rapproche de l'ambiante, le
terme en q* , se comportant d'après (22b) comme la densité à la
puissance 2/3, devant peu varier et légèrement augmenter quand la
température baisse.

ln(6/Hz)
1Oi

A. à.

2 2.4 2.8 3.2 3.6

1/T(UT3K'1)

figure 15 : Variation du paramètre d'élargissement 6 du PTFE pur en

fonction de la température.
'}
 164

-38

-60

-128

figure 16 : Formes de raie du déplacement chimique du PTFE amorphe en

fonction de T [32].

L'analyse des spectres RMN de polymères ne présentant

macroscopiquement pas de phase nématique peut donc s'inspirer du
traitement des données de RMN des nématiques.
En considérant que, localement, l'ordre au sein de ces polymères
est analogue à celui au sein d'une phase nématique, la différence ne
se faisant qu'au niveau de la corrélation orientationnelle qui, si
elle s'étend sur tout l'échantillon dans les cristaux liquides, ne
s'étend ici que sur un monodomaine réduit (dont la taille reste à
déterminer), on peut reproduire de manière satisfaisante leurs
spectres en phase amorphe qui s'avèrent eux aussi soumis à l'effet des
modes élastiques. Cette analyse, assez fastidieuse, ne peut néanmoins
ï se faire ex nihilo, un spectre de poudre contenant bien moins
d'informations qu'un spectre aligné, mais peut apporter des précisions
tant sur la structure ou les mouvements internes à la molécule que sur
les propriétés visco-élastiques du matériau.

t
J
 CONCLUSION

'

?
\ 1.
 167

Cette étude a permis d'étendre le champ d'action du modèle à une

conformation à de nouveaux nématiques de faible masse moléculaire
ainsi qu'à une série de polymères, certains ne présentant pas de phase
nématique dans leur phase pure. En partant d'une molécule de base dont
on peut calculer le spectre théorique en supposant qu'elle possède le
même nombre de conformations dans les phases solides et nématiques et
en ne considérant que les mouvements internes à cette molécule, ce
spectre théorique doit ensuite subir les effets des mouvements
externes à la molécule, qui s'ils sont suffisamment rapides peuvent
être traduits simplement par un unique tenseur d'ordre mais
nécessitent l'utilisation d'un formalisme plus sophistiqué, tel celui
décrit dans le chapitre T, dans le cas où des modes élastiques lents
peuvent influencer la forme de raie, comme c'est le cas pour de
nombreux polymères.
Un premier exemple a été trouvé, en ce qui concerne les
nématiques de faible masse moléculaire où l'influence des modes
élastiques lents peut être négligée, avec le 5CB pour lequel
l'application du modèle à conformation unique ne donne pas de résultat
satisfaisant, du moins s'il est pris dans son sens littéral. Cet échec
apparent a permis de prédire l'existence d'un désordre en bout de
chaîne dans le 5CB en phase nématique, existence par ailleurs
fortement supportée par un phénomène semblable dans la phase solide de
la molécule proche de 4CB..
D'autre part, le modèle à conformation unique a démontré son
intérêt dans l'étude des spectres RHN de polymères nématiques (les
ME9Sn) et également de polymères conventionnels tels le PTFC et le
PTFE, même si dans ces deux derniers cas des incertitudes persistent.
Considérer que l'ordre local, dans les phases nématiques et
amorphes, est semblable à celui observé dans la phase solide et qu'on
peut y définir une sorte de cellule unité, notre molécule de base,
laquelle est soumise à des mouvements collectifs telles les
fluctuations visco-élastiques, apparaît comme une description
prometteuse permettant d'analyser les spectres de polymères. Cette
analyse peut alors fournir des informations sur la conformation des
molécules (unique dans la plupart des cas), leurs mouvements internes
(opérations de symétrie au sein de la molécule de base) et les
f. mouvements collectifs (liés ici aux propriétés visco-élastiques).
 169

REFERENCES BIBLIOGRAPHIQUES

[I] O.W. Doane dans Magnetic Resonance of Phase Transitions, édité

par F.J. Owens, C.J. Poole et H.A. Farach (Academic Press, 1979),
Chap. 4, 171.

[2] O.W. Emsley et J.C. Lindon, NHR Spectroscopy using Liquid

Crystals Solvents (Pergamon Press, 1975).

[3] The Molecular Physics of Liquid Crystals, édité par G.R.

Luckhurst et G.W. Gray (Academic Press, 1979).

[4] P. Diehl, dans NMR of Liquid Crystals, édité par J.W. Emsley (D.
Reidel, 1985), 147.

[5] a) S. Grande, St. Limmer, H. Schmiedel et R. Stannarius, dans

j Selected Topics in Liquid Crystals Research, édité par H.D.
Koswig (Akademie-Verlag, Berlin, 1990).
b) St. Limmer, Fortschr. Phys., 37, 879 (1989).

[6] F. Volino et D. Galland, MoI. Cryst. Liq. Cryst., 212, 77 (1992).

[7] D. Galland et F. Volino, J. Phys. France E, 1, 209 (1991).

[8] a) J.B. Ferreira, A.F. Martins, D. Galland et F. Volino, MoI.

Cryst. Liq. Cryst., 151. 319 (1987).
b) D. Galland et.F. Volino, J. Phys. France, 50, 1743 (1989).
c) F. Volino, D. Galland, J.B. Ferreira et A.J. Dianoux, Journal
of Molecular Liquids, 43, 215 (1989).

[9] P. Esnault, J.P. Casquilho et F. Volino, Liquid Crystals, 3, 1425

(1988).

[10] P.G. de Gennes, The Physics of Liquid Crystals (Clarendon Press,

1974).

[II] F. Reinitzer, Wiener Monatsh., £, 421 (1888).

"•{
'J [12] F.M. Leslie, Arch. Rat. Mech. Anal., 28, 265 (1968).

[13] O. Parodi, J. Physique, 31, 580 (1970).

J [14] M. Miesowicz, Nature, 158, 27 (1946).

 170

* t [15] a) Orsay Liquid Crystal Group, J. chem. Phys., 51» 816 (1969).
b) H.J. Coles and M.S. Sefton, MoI. Cryst. Liq. Cryst. Lett., I,
151 (1985).
!
[16] A. Abragam, The Principles of Magnetic Resonance (Oxford
University Press, 1960).

[17] O. Doucet, J.P. Mornon, R. Chevalier et A. Lifschitz, Acta

Cryst., B33, 1701 (1977).

[18] F. Volino, A.J. Dianoux, R.E. Lechner et H. Hervet, J. Phys.

France, Colloques 36, Cl-83 (1975).
[19] M. Warner, Holec. Phys., 52, 677 (1984).

[20] J.B. Ferreira, H. Gérard, D. Galland et F. Volino, Liquid

Crystals, 13, 645 (1993), joint à ce texte.

[21] H. Gérard, J. Avalos, D. Galland et F. Volino, Liquid Crystals,

' 12, 649 (1992), joint à ce texte.

[22] D. Catalane, L. di Bari, C. A. Veracini, G. N. Shilstone and C.

Zannoni, J. chem. Phys., 94> 3928 (1991).

[23] H. Gérard, D. Galland et F. Volino, non encore publié, joint à ce

texte.

[24] D. Galland, H. Gérard, J.A. Ratto, F. Volino et O.B. Ferreira,

Macromolecules, 25, 4519 (1992), joint à ce texte.

[25] J.A. Ratto, F. Volino et R.B. Blumstein, Macromolécules, 24. 2862

(1991).

[26] P. Coulter et A.H. Windle, Macromolecules, 22, 1129 (1989).

[27] Handbook of Physics and Chemistry, 65e"16 édition (CRC Press,

1984)
5|%
J
\ [28] M. Mehring, R.G. Griffin et J.S. Waugh, J. chem. Phys., 55, 746
(1971).
* i
* [29] E.S. Clark et L.T. Muus, Z. KHst., 117. 119 (1962).

[30] A.J. Gordon et R.A. Ford, The Chemist's Companion (Wiley, 1972).

:.f [31] A.N. Garroway, D.C. Stalker et P. Mansfield, Polymer, 16, 171
J (1975).
 .ai
171

[32] A.J Vega et A.D. English, Macromolecules, 13, 1635 (1980).

vl

•f-
 y

.at.

*
•I
vl

<*>'
œa
CENG / SERVICES TECHNIQUES / REPROGRAPHIE / 1 993
Nom de l'auteur : Hervé GERARD

Etablissement : Université Joseph Fourier (Grenoble I)

Le but de cette étude est la simulation et l'exploitation de spectres de RMN de cristaux liquides
nématiques et de polymères. Les formes de raie RMN sont analysées grâce à deux modèles
complémentaires, le premier (modèle à conformation unique) décrivant la contribution purement
moléculaire (géométrie et mouvements internes à la molécule), le second la contribution des
mouvements collectifs (modes visco-élàstiques). Des rappels sur la méthode RMN et la notion d'ordre
orientationnel au sein de la phase nématique sont fournis dans la première partie où sont également
décrits ces deux modèles. Dans une deuxième partie, ces modèles sont appliqués aux données
relatives à des molécules nématiques de faible masse moléculaire ainsi qu'à des polymères
nématiques. Cette application permet d'obtenir des informations sur la structure et les mouvements
internes à la molécule. Tordre orientationnel régnant au sein de la phase et les propriétés visco-
élastiques du matériau étudié. Enfin, on montre que l'extension de ces modèles aux données RMN de
polymères ne présentant pas de phase nématique en phase pure permet d'obtenir des informations
similaires en considérant que leur phase amorphe présente localement un ordre nématique.

Mots clés:
cristaux liquides nématiques, polymères nématiques, polymères, RMN,
simulation de formes de raie, constantes visco-élastiques, ordre orientationnel.