You are on page 1of 34

Small-Angle X-Ray Scattering Studies of

Microdomain Structure in Segmented Polyurethane


Elastomers*

JEFFREY T. KOBERSTEIN,* Polymer Materials Program, Department of


Chemical Engineering, Princeton University, Princeton, New Jersey 08544
and RICHARD S. STEIN, Polymer Research Institute, University of
Massachusetts, Amherst, Massachusetts 01003

Synopsis
The small-angle x-ray scattering (SAXS) technique has been used to characterize the detailed
microphase structure of two crosslinked segmented polyurethane elastomers. Both copolymers
contain trifunctional polypropylene ether triols in the rubbery elastomeric block, but are synthesized
with different hard segments: a symmetric 4,4’-diphenylmethane diisocyanate (MDI) chain extended
with butanediol (BD); and an 80/20 mixture of asymmetric 2,4-toluene diisocyanate (TDI) and
symmetric 2,6-toluene diisocyanate (TDI) chain extended with ethylene glycol (EG). Calculations
of SAXS invariants and determinations of deviations from Porod’s law are used to examine the degree
of phase segregation of the hard- and soft-segment domains. Results show that the overall degree
of phase separation is poorer in the asymmetric TDI/EG-based copolymer than in the symmetric
MDI/BD-based copolymer. Determination of diffuse phase boundary thicknesses, however, reveals
that the domain boundaries are sharper in the asymmetric TDI/EG system. The contrasting mor-
phologies found in the two systems are interpreted in terms of differences in hard-soft segment
compatibility, diisocyanate symmetry, and diisocyanate length. Coupled with conformational
considerations, this information is used to construct a new model for polyurethane hard-segment
microdomain structure. Important features of the model are that it takes into account the effects
of hard-segment sequence length distribution and allows for folding of the longer hard-segment se-
quences back into the hard-segment domain.

INTRODUCTION
Segmented polyurethane elastomers are thermoplastic (AB), -type segmented
copolymers that find extensive industrial applications as a result of mechanical
behavior that is similar to conventional crosslinked rubbers. The unusual
elastomeric properties of these materials have been attributed to the formation
of a microphase separated domain structure consisting of “hard”-segment-rich
and “soft”-segment-rich domains. The soft-segment domains are rubbery a t
service temperatures, while the hard-segment domains are glassy. The latter
are thought to act as thermally labile physical crosslink sites and as filler material.
This thermal lability allows polyurethanes to be processed as thermoplastic
materials, yet yields materials with the mechanical behavior of thermosets.
The consequently strong industrial interest in polyurethane copolymers has
spurred a great number of investigations into their structure-property correla-
tions. A comprehensive review of these studies is not presented here, but will

* To whom correspondence should be addressed.

Journal of Polymer Science: Polymer Physics Edition, Vol. 21,1439-1472 (1983)


01983John Wiley & Sons, Inc. CCC 0098-1273/83/081439-32$04.20
1440 KOBERSTEIN AND STEIN

be the subject of a future publication. As yet no comprehensive understanding


of polyurethane structure has emerged, owing for the most part to the complexity
of the materials themselves. The synthesis of segmented polyurethane co-
polymers involves simultaneous condensation copolymerization of three pre-
cursory compounds: intermediate-molecular-weight polyol, diisocyanate, and
chain extender. There are a multitude of precursor materials available to choose
from so that the possible number of polyurethane chemical structures obtainable
is very large. Each of these structures will have associated with it a particular
morphology and mechanical response. The task of formulating structure-
property relations is thus a formidable one.
This task is further complicated by the extensive physical and thermodynamic
interactions that are manifested in polyurethanes. It is well known that the
mechanical properties of these materials are highly dependent on the degree of
microphase segregation. The degree of phase separation, however, is not only
a function of the system thermodynamics, but also depends a great deal on the
hydrogen bonding characteristics and ability for hard segments to pack correctly
to form hydrogen bonds. The propensity for hard-segment packing is in turn
directly related back to the chemical structure and composition of the copolymer.
Structural factors that may affect hard-segment packing include diisocyanate
size and symmetry; chain extender length and functionality; and polyol type,
molecular weight, and functionality. In addition, there is a distribution of
hard-segment lengths, which is highly dependent on the stoichiometry and
conversion of the condensation copolymerization. Finally, hard-segment packing
is often a nonequilibrium process, which leads to a dependence of material re-
sponse on the thermal and mechanical history of the sample.
A complete understanding of these complicated materials therefore requires
careful consideration of all of these chemical and physical factors. The aim of
the investigations undertaken in this report is deeper understanding of poly-
urethane structure-property correlations through characterization of the mi-
crodomain structure in polyurethanes prepared from two different diisocy-
anates.
The report examines carefully the effect of intradomain ordering of hard
segments on the overall copolymer morphology, through comparison of systems
based on asymmetric and symmetric diisocyanate precursors. Hard segments
prepared from the symmetric diisocyanate are capable of well-developed para-
crystalline or even truly crystalline order. Hard segments based on asymmetric
diisocyanates, on the other hand, do not generally crystallize, and show at best
a paracrystalline ordering. The detailed domain morphology of these materials
is characterized by small-angle x-ray scattering (SAXS) analysis in order to as-

TABLE I
Chemical ComDosition of Polyurethanes
Hard Soft Monomer
Designation segment segment 4HS ratio
MDI/BD MDI-BD PPET" 0.24 9:81
TDI/EG TDI-EG PPED-PPETb 0.21 10:9:1
a 10,000-MW polypropylene ether triol.
50/50 mixture of 8300-MW polypropylene ether dioU7475-MW polypropylene ether triol.
SMALL-ANGLE X-RAY SCATTERING STUDIES 1441

certain what effect this intradomain ordering and hard segment structure may
have on the material structure and associated properties. A model for poly-
urethane domain structure, based on these results is proposed. The model
provides reasonable explanations for the effects of hard-soft segment compati-
bility, diisocyanate symmetry, and diisocyanate length on polyurethane mor-
phology.

EXPERIMENTAL

Samples
The polyurethane block copolymers were kindly prepared by Dr. Robert
Herold of the General Tire and Rubber Company of Akron, OH. Chemical
compositions of the samples are shown in Table I. The material denoted
MDI/BD is based on a symmetric hard segment consisting of 4,4'-diphenyl-
methane diisocyanate (MDI) chain extended with butane diol (BD), and a 10,000
(number-average) molecular weight soft segment of polypropylene ether triol
end capped with 10 w t % of ethylene oxide. The sample denoted TDI/EG con-
tains hard segments formed from toluene diisocyanate (TDI) and ethylene glycol
(EG). The TDI used is an 80:20 commercial mixture of the asymmetric 2,4
isomer and the symmetric 2,6 isomer. The soft segment for this material is a
5050 mixture of 8,300 number-average molecular weight polypropylene ether
diol and 7,450 molecular weight polypropylene ether triol. Schematics of the
chemical structures of these two materials are drawn in Figures 1 and 2. The
two samples have similar hard-segment content and calculated hard-segment
sequence length distribution as shown in Table I.
The polyurethanes were prepared by the two-step or prepolymer procedure.
Diisocyanate end-capped prepolymers were first prepared by reacting the polyol
prepolymer with a slight excess of diisocyanate and subsequently adding the
chain extender. The prepolymer was then mixed with the remaining isocyanate
and a small amount (ca. 0.02 wt %) of dibutyltin dilaurate catalyst. The resulting
syrup was degassed, poured into the desired mold, and cured at 50°C for one hour

MDI /ED SYSTEM DIISOCYANATE

O H HF!
- (a),-,
U UGU) ,r u =+ ~ - N * H ~ ~ N - c +

CHAIN EXTENDER

G= O-(CH,),- 0 -f

U(GU),T
s(xT SEGMENT

9
&C-(CH,-CH-O)~(CH,Cb&Ot,

YI HC-("

H,C- ( 'I
")-("

")-('I

Fig. 1. Chemical structure schematic of the symmetric 4,4'-diphenylmethane diisocyanate


"

'I
1-
)-

(MD1)-based segmented polyurethane block copolymer.


1442 KOBERSTEIN AND STEIN

TDVEG SYSTEM DII SOCYANATE


PY
U = +C-N
a CH,

CHAIN EXTENDER

U G = +OCH,CbO+

SOFT SEGMENTS

Fig. 2. Chemical structure schematic of the asymmetric toluene diisocyanate (TD1)-based seg-
mented polyurethane b!ock copolymer.

plus 70°C for 18 h. Throughout the polymerization procedure, care was taken
to exclude atmospheric water from the reaction mixture. All experimental
measurements were performed on the as-polymerized samples.

Thermal Analysis
Thermal analysis was carried out on a Model 1B Perkin-Elmer differential
scanning calorimeter (DSC). Samples of 10-15 mg mass were run under inert
gas at 10°C/min at a sensitivity range setting of 4. Materials were characterized
as received over a temperature range beginning a t room temperature and ter-
minating when melting or degradation was observed. Measurements of glass
transitions were carried out by precooling with liquid nitrogen.

Dynamic Mechanical Analysis


The dynamic mechanical response was evaluated with a Rheovibron DDV-IIB
dynamic viscoelastometer (Toyo Baldwin Co.). Samples were tested a t 3.5 Hz
over a temperature range of approximately -100 to 75°C under dry nitrogen
atmosphere. Data analysis was accomplished with the aid of a computer program
furnished by Senich.l

SAXS
SAXS measurements were recorded using two different apparatuses, each with
a different scattering geometry and collimation system. The majority of the
work was performed on a Kratky camera aligned for infinite slit optics. Com-
plementary measurements, using pinhole collimation, were made a t the 10-m
SAXS facility at Oak Ridge National Laboratories.2 The Oak Ridge facility uses
SMALL-ANGLE X-RAY SCATTERING STUDIES 1443

a two-dimensional position-sensitive detector, while the other camera was


equipped with a step-scanning proportional counter.
Profiles were corrected for parasitic scattering with the sample placed directly
on the detector face as a more accurate means of attenuating the background
scattering. Raw scattering profiles were computer smoothed by using a pro-
gressive piecewise cubic-fit smoothing routine. This was carried out to minimize
errors in desmearing procedures and calculation of correlation functions, and
also facilitates analysis of data in the Porod-law region, where statistical scatter
is large. These procedures were carried out by using a modification of Vonk’s
program H10g3 in which a Lupolen standard supplied by Kratky is employed
for intensity calibration. Samples for SAXS measurements were polymerized
directly in the form of 2-mm-thick plaques.

SAXS ANALYSIS
SAXS is an extremely useful tool for the characterization of microphase
structure in polymeric materials. Careful application of this analysis provides
measurements of interdomain spacings, domain boundary diffuseness, and the
degree of microphase separation. Since this study relies heavily on detailed
SAXS analyses we present a fairly comprehensive review of the appropriate
scattering theory. This will facilitate subsequent discussion of experimental
results.
The simplest of these analyses involves the application of Bragg’s equation
to determine interdomain spacings; i.e., the position of a maximum in the scat-
tering profile is related to the interdomain spacing d by
(sI = (2 sinO,)/X = n/d (1)
where X is the wavelength, n is the order of reflection, 28, is the angular position
of the maximum in the scattering profile, and IsI is the magnitude of the recip-
rocal lattice vector. For simplicity the scattering relations that follow are written
in terms of the scattering vector h instead of s. The two are related, however,
by the relation I hl = (47r sinB)/X = 2 ~ 1 ~where
1 , 28 is the scattering angle.
Interdomain spacings may also be obtained through the correlation function -
approach. For spherically symmetric systems this approach yields a relation
for the Rayleigh factor of scattered radiation of the form4

The Ftayleigh factor is independent of the scattering apparatus and depends only
on the sample used. It is defined as

where I ( h )is the scattered irradiance corresponding to a scattering vector h, p


is the sample to detector distance, I0 is the incident beam irradiance, and V is
the illuminated sample volume. The term i, is the Thomson scattering factor
for a single electron.
The three-dimensional correlation function y3D(r) contains all the information
pertaining to the shape and spatial distribution of the phases giving rise to the
scattering. This function expresses spatial correlations between local fluctua-
1444 KOBERSTEIN AND STEIN

tions in electron density, and is defined as


YYD(r) = ( f b j A p k (4)
where Apj is the local fluctuation of the electron density from the average, the
brackets refer to an average over all fluctuations separated by a constant r, and
-
Ap2 is the mean-square fluctuation. The value of the correlation function is
related to the probability that a rod of length r will have both of its ends in the
same phase. The limiting values are therefore unity when r = 0 and zero as r
approaches infinity. A maximum in the correlation function is indicative of
periodicity in phase correlations and may be used to characterize interdomain
spacings.
The experimental three-dimensional correlation function may be obtained
by taking the Fourier transform of the experimental scattering profile
1 m

Y3D(r) = dh
27r2i, Ap2
If a lamellar morphology for the microdomain structure is assumed, the SAXS
analysis must be modified. The amplitude of scattered radiation for a one-
dimensional lamellar system is5

where the one-dimensional correlation function is defined as

ylD(r) = -
+
Ap ( x r) Ap(x )dx
(7)
AP2
In reality, however, the scattering profile is isotropic. That is, the lamellae have
limited coherence and can be thought of as lamellar bundles oriented equally
in all directions. The scattering profile then contains information relating to
distance correlations averaged over all directions. Information pertaining to
spatial correlations perpendicular to the lamellae may be obtained by realizing
that the scattered intensity is spread out equally about the surface area of spheres
in reciprocal space. The real intensity I ( h )is then related to the one-dimensional
intensity i(h)by
Z(h) = i(h)l4~1s)~ (8)
As a consequence of this effect, the true Bragg spacing would be found by
taking the inverse of the value of Is1 at which a maximum in the function h2B(h)
[or s 2 3 ( s ) ]is found. Similarly, the experimental one-dimensional correlation
function may be calculated from the following relation:

Jm h2R(h)cos(hr)dh
ylD(r) = (9)
Jm h2R(h)dh

In order to determine the experimental correlation function according to (5)


or (9),it is necessary to extrapolate the scattering profile to infinite scattering
vector. This may be accomplished through Porod's law. For two-phase
SMALL-ANGLE X-RAY SCATTERING STUDIES 1445

a) I
X-

-I
Ad

b) 1
X-

X-

Fig. 3. Characteristicelectron density profiles: (a) real system, (b) corrected for thermal density
fluctuations, (c) corrected for thermal density fluctuations and diffuse phase boundaries.

structures with sharp phase boundaries, corresponding to the electron density


profile shown in Figure 3(c), this is given by6
K
lim [R(h)]= -
h-- h4
where K is a constant related to the surface to volume ratio of the phases.
Real systems, however, contain two structural features which lead to deviations
from this law. Diffuse phase boundaries provide a profile as depicted in Figure
3(b). The corresponding Porod's-law relation is7
KH2(h)
lim [R(h)]= ~

h-m h4
where H(h) is the Fourier transform of a function which accounts for smoothing
of the phase boundaries. If the system has a sigmoidal gradient in the phase
boundary [Fig. 4(a)], the Fourier transform of the smoothing function is given
by
H(h) = exp(-'/202h2) (12)
The value of c is related to the standard deviation of a Gaussian function used
to generate the diffuse boundary and thus serves as a measure of its thickness.
If the diffuse boundary gradient is linear with thickness E [Fig. 4(b)],the Fourier
transform of the smoothing function becomes8
H ( h ) = [sin(Eh/2)]/(Eh/2) (13)
1446 KOBERSTEIN AND STEIN

0.606

-3 -2 -I 0 I
r-e
(a 1

I
SMOOTHING I,E
FUNCTION
I
Fig. 4. Diffuse-boundary models: (a) sigmoidal gradient, (b) linear gradient. From ref. 12.

In addition to the presence of diffuse boundaries that give rise to “negative”


deviations from Porod’s law, the actual electron density profile contains fluc-
tuations due to thermal density effects. This contribution gives rise to a back-
ground term W B( h ) ,leading to “positive” deviations from Porod’s law.9 The
Porod’s-law relation corresponding to the electron density profile of the “real”
scattering system [Fig. 3(a)] is then

h-a
~

h4
+
lim [W(h)] = K H 2 ( h ) W,(h)

The background term generally has a slowly varying dependence on h , and may
be assumed to be a constant, FI. The contribution of this term may then be
accounted for by making an experimental plot of W(h)h4vs. h4.7J0 From (14)
it is seen that this plot is linear with slope W,(h) = FZ.
The complete procedure for extrapolation thus involves determining of RB(h),
subtracting this contribution, and then fitting (11) to the limiting scattering
intensity. The latter procedure yields a value of the diffuse boundary thickness
for one of the models shown in Figure 4. The functional form of the scattering
profile in the asymptotic limit is then completely specified by (14),from which
extrapolations beyond the limits of measurable data can be made. With these
extrapolations, the experimental correlation functions may be determined
through application of (5) and (9).
The mean-square variance in electron density, G2, may be calculated through
SMALL-ANGLE X-RAY SCATTERING STUDIES 1447

determination of the Porod invariant?

The experimental determination of the invariant also requires extrapolation of


measured scattering profiles so that this integral may be calculated. This is
accomplished through application of the Porod's-law relations, as was done for
calculation of correlation functions.
The electron density variance serves as a measure of the degree of phase mixing
in a system. For an ideal two-phase system, the variance in electron density
is
-2
APC = 4142(Pl - P2I2
where $1 and p 1 are the volume fraction and electron density of the ith phase,
respectively. This value may be calculated from the known chemical composi-
tion and densities of the component phases by assuming complete phase sepa-
ration.
The experimental electron density variance is less than this ideal value as a
result of the presence of mixing within phases and diffuse boundaries as depicted
in Figure 3. Experimental electron density variances may be calculated for these
profiles by employing the Porod's-law relations developed earlier.1° It was
previously shown that thermal density fluctuations give rise to a scattering
background B E ( h )which may be determined experimentally. On removal of
this contribution, the electron density variance G2', corresponding to the model
shown in Figure 3(b), is defined as

It has also been shown that the effects of diffuse phase boundaries on the scat-
tered intensity are calculable. After correcting the intensity for these effects,
the idealized profile of Figure 3(c) is obtained. The corresponding variance G2"
may be calculated from

where H ( h ) is the Fourier trans€orm of the smoothing function.


The comparison of these three model variances (Gz, z2', and G2") provides
valuable inferences into the state and degree of phase separation, as shown by
Bonart et al.lOJ1 The degree of overall phase separation is reflected in the
ratio

which assumes a value of unity if the system has achieved complete phase seg-
regation and decreases with the occurrence of phase mixing. The quantity
Zpy&P - 1
is indicative of the amount of phase mixing within domains, irrespective of dif-
fuseness of phase boundaries. It is zero in the absence of this type of mixing,
and increases with the occurrence of intradomain segmental mixing. The
1448 KOBERSTEIN AND STEIN

presence of diffuse boundary mixing, irrespective of intradomain mixing, is in-


dicated by

which is zero for systems with sharp phase boundaries and increases with in-
creasing domain boundary diffuseness. The information obtained in this manner
is semiquantitative, but is extremely valuable in assessing the degree and origin
(within phases or a t the phase boundary) of phase mixing.
Direct interpretation of the electron density variance in terms of a two-phase
model [eq. (16)]is not possible for systems with mixing within domains. This
is due to the fact that the volume fractions and electron densities are not inde-
pendent in these systems. However, the effect of mixing a t domain boundaries
may be taken into account by using a modification of eq. (16). With the linear
gradient [Fig. 4(b)] between phases, for example, the electron variance may be
written8
-
-I~
Ap2’ = (4142 - E S I ~ V ) ( P 2 ) ~ (22)
where SIV is the specific surface of the phases. It is not possible to derive the
corresponding relation for the sigmoidal gradient case. This is due to the fact
that the Gaussian smoothing function used to generate this electron density
profile does not have finite limits. The “diffuseness” of the boundary for this
model actually persists to infinity, making derivation of such a relation impossible
since a domain is of finite dimension. In this case the method described by (17)
must be used. In principle the methods described by (17) and (18)are preferred
since they can be applied without assuming, a priori, any model for the diffuse
boundary gradient. This can be accomplished by forcing the scattering curve
to obey Porod’s law, from which H2(h)is obtained.
T o the authors’ knowledge, this process has never actually been carried out,
and it is not clear how feasible it would be. General practice is to determine
H2(h)through fitting of the Porod’s-law relation [eq. (14)]for a particular model.
When H2(h)is known, the variance is calculated from (18).
The scattering theory presented in the preceding discussion applies to scat-
tering systems in which a pinhole collimating system is employed. It is often,
however, advantageous to use a slit collimation system, such as is present in a
Kratky camera. This scattering geometry leads to a smearing of the scattering
curve. There are two manners in which this effect may be accounted for. The
experimental scattering curve may be desmeared by various numerical proce-
dures, or the theory may be smeared analytically.
If the slits may be assumed to be of infinite length, as in the Kratky system,
the desmeared intensity R(h)may be calculated from the smeared intensity %(h)
through the following relation5

After this integral is evaluated numerically from experimental data, the pinhole
intensity R ( h )is obtained. This procedure is necessary in order to calculate
a Bragg spacing or to determine the experimental correlation functions from the
smeared scattering data. The desmearing procedure, however, introduces error
SMALL-ANGLE X-RAY SCATTERING STUDIES 1449

into the analysis due to a magnification of statistical errors or due to truncation


effects involved with the numerical evaluation of the integral in (23).
These errors are largest in the Porod's-law region, where the scattered intensity
is low and the statistical error is high. For these reasons it is preferable to develop
smeared scattering relations for analysis of deviations from Porod's law and
calculations of invariants. For the case of infinite-length slits of negligible width,
the smeared scattering intensity is given bys

% ( h )= 2C J m R ( h 2 + u2)lI2du
where C is a determinable constant. From this relation, smeared forms for
scattering relations may be derived. A smeared form of Porod's law may be
defined as
TCK
lim%(h) = -
h-m h3

Similarly a relation for deviations from Porod's law may be written as15

h+m

where A(h)is the smeared form of the Fourier transform of the smoothing
function and %B ( h )is the smeared background intensity. If the background
is constant, this latter contribution may be evaluated through a plot of %(h)h3
vs. h3.l0 The slope of such a plot gives the constant background. The smeared
form of the Fourier transform of the smoothing function, B ( S )is, determined
by substituting (14) into (24) and integrating. The exact solution to this integral
is known for the sigmoidal interface gradient model,1° but it is not amenable to
graphical analysis. Approximate solutions for both the linear and sigmoidal
gradient models have been presented in a recent review.12 For the case of the
sigmoidal gradient it was shown that
A 2 ( h )= e ~ p [ - 1 . 3 6 ( a h ) l . ~ ~ ] (27)
is a good approximation to the exact solution. The diffuse boundary thickness
parameter CT is determined from a plot of ln([g(h)- % ~ ( h ) ]vs. ~ ] A value
h hl.sl.
of the width E of a linear-gradient diffuse boundary is obtained by realizing that
E =n u .
Finally, the electron density variances may be evaluated directly from smeared
intensity profiles through the following relationslOJ1:

These relations provide a means of characterizing the state of phase separation


[through application of (19-21)] directly from smeared scattering data. This
is accomplished in a manner similar to that described in the text following eq.
(21).
1450 KOBERSTEIN AND STEIN

ii
I
I-

z
w
a
0

I
1 I

To provide completeness as well as serve as a check of internal consistency the


SAXS analyses have been carefully performed on both desmeared and smeared
data as well as data taken with pinhole collimation. A detailed discussion of the
analyses for diffuse boundary thicknesses is the subject of a paper to follow.

RESULTS AND DISCUSSION

Thermal Behavior
DSC measurements on the TDI/EG copolymer do not reveal any significant
endothermic features. The high-temperature thermal response curve shows
only a gradual change in heat capacity. This is consistent with the results of
Schneider et al. for similar copolymers.l3
The endothermic response of the MDI/BD copolymer (Fig. 5) is rich in com-
parison. Three distinct regions of activity are identifiable in this curve, corre-
sponding well to those found by DTA and DSC studies of other MDI-based co-
polymers.14-'6 In addition, the subambient DSC thermograms show a glass
transition for the copolymer over the temperature range -68 to -63°C.
The 70°C endotherm is barely discernible in this copolymer, a result which
is probably related to the long hard-segment sequence length. Similar endo-
therms have been attributed to the dissociation of urethane soft-segment hy-
drogen bonding and have been found experimentally to decrease in intensity with
increase in hard-segment sequence length.15 This interpretation is consistent
with thermodynamic arguments. An increase in sequence length results in
improved phase separation, and thus a decrease in the number of urethane
soft-segment contacts.
Two additional endotherms are observed a t 180 and 215°C. The postulated
origin of these endotherms is the dissociation of interurethane hydrogen bonding
within ordered hard-segment domains. Wide-angle x-ray scattering profiles
show that the order is not truly crystalline. The sole features observed are an
amorphous halo accompanied by a very weak reflection a t ca. 8 A-1. Similar
reflections have been observed by Bonart et al.17J8and others,l9 and have been
attributed to formation of pseudolattice planes in the hard-segment domains.
SMALL-ANGLE X-RAY SCATTERING STUDIES 1451

Fig. 6. Smeared relative small-angle x-ray scattering profiles for MDI/BD (open circles) and
TDI/EG (open triangles).

These reflecting lattice planes are ascribed to preferential but noncrystalline


alignment of hard segments.
Endotherms in the 215°C region are generally attributed to regions that may
be microcrystalline but perhaps are not large enough to give rise to crystalline
x-ray reflections. The melting temperature of pure hard segments is reported
by MacKnight et al. to be 248"C.20 This is in contrast to the value of 208°C re-
ported recently by Camberlin et aL21for 1022-MWhard segments. The apparent
discrepancy may be due to molecular weight differences. MacKnight et al. do
not report the molecular weight of their sample. The average molecular weight
of the MDI/BD copolymer studied in this work is ca. 3,000. In light of these T,
values it is reasonable then to assign the 215°C endotherm to microcrystalline
ordered regions within hard-segment domains. The 180°C endotherm can be
understood in terms of the presence of regions within hard-segment domains
where order is not microcrystallinebut rather paracrystalline or liquid crystalline.
Disorder leading to paracrystallinity may be justified by considering the effects
of sequence length distribution on chain packing.
The absence of true crystallinity in this material (with long sequence lengths)
is conceivably due to the thermal history of the material. The material was bulk
polymerized at 70"C, which may be close to the hard-segment Tgand not suffi-
ciently high to permit significant crystallization.
Although the experiments were not carried out, it is reasonable to believe that
1452 KOBERSTEIN AND STEIN

I I

s (nm-')
Fig. 7. Desmeared small-angle x-ray scattering profiles for MDIBD (open circles) and TDI/EG
(open triangles).

the positions of the observed endotherms are dependent on thermal history and
that microcrystallinity can be promoted by proper thermal annealing.19~~~.~3 In
this regard the interpretation of the DSC endotherms for the-MDI/BD material
is that they are of morphological origin, arising from the presence of paracrys-
talline or perhaps liquid-crystalline order.
The corresponding results for the TDI/EG system demonstrate that asym-
metry of the dissocyanate disrupts significantlythe ordering within hard-segment
us
domains. This is an important res t to keep in mind as it is central to the in-
terpretation of the SAXS data and subsequent construction of the model for
microphase structure.

SAXS Measurements
The Kratky slit-smeared small-angle x-ray scattering profiles for the two
polyurethanes appear in Figure 6. Both curves show a single maximum followed
by a gradual decrease in intensity. The peak is better-defined and comes at
larger scattering angle for the TDI/EG polyurethane. The decrease in intensity
is also steeper for these materials. The same trends are seen in Figures 7 and
8, which show the collimation-corrected profile and the collimation-corrected
lamellar model profile, respectively. As a test of the validity of the collimation
correction, scattering profiles taken with pinhole collimation are compared tot
the slit-collimated corrected profiles in Figures 9 and 10. Both plots show good
agreement, except that maxima are broader for pinhole curves, and that devia-
tions are more pronounced a t smaller s values. These discrepancies are con-
sistent with the slit smearing effect caused by the finite size (0.1 X 0.1 cm) of the
pinhole. It is not obvious why the discrepancies are stronger for TDI/EG;
however, it may be related to the sharpness of the slit-smeared scattering max-
imum in this material. The desmearing procedures employs differentiation of
SMALL-ANGLE X-RAY SCATTERING STUDIES 1453

I I
Ix s2
I10
2i

0
0 .O5 .I0 15
s (nm")
Fig. 8. Desmeared small-angle x-ray scattering profiles corrected for lamellar geometry;MDIDD
(open circles), TDI/EG (open triangles).

a Fourier-series curve fitting, which is necessarily less precise for a sharper


maximum. The discrepancies are not believed to be important, however, as they
appear in a region of the scattering curve that is not analyzed in detail.

A A

A A

A
A
A
AA
A

oh O.b2 O.b4 0.66 o.b* 0.lo 0.


S ( nm-l)
Fig. 9. Comparison of small-angle x-ray scattering profiles for TDIEG; pinhole collimation (open
triangles); desmeared slit collimation (filled triangles).
1454 KOBERSTEIN AND STEIN

4 0102 oh4 0.k o.bs o.:o 0.12

Fig. 10. Comparison of small-angle x-ray scattering profiles for MDIDD; pinhole collimation
(open circles), desmeared slit collimation (filled circles).

Interdomain Spacings
Interdomain spacings or d spacings calculated from peak positions using
Braggs law are listed in Table 11. The one-dimensional column refers to the
position of the maximum in scattering profiles corrected for the lamellar model
(Zd,,,s2) as depicted in Figure 8. Values of interdomain spacings were also de-
termined by correlation function analysis. The analysis procedure was carried
out on scattering profiles corrected for the effects of diffuse phase boundaries
and thermal density fluctuations. The correlation function thus represents
spatial correlations between domains only. The three-dimensional correlation
functions are shown in Figure 11. A maximum corresponding to a spacing of
14.5 nm is clearly seen for the TDI/EG polyurethane. The MDI/BD material
shows only a very weak maximum a t ca. 21 nm. Correlation functions assuming

TABLE I1
Interdomain Spacings-Kratky Desmeared" (nm)
Three-dimensional One-dimensional
Sample d [from 3D y ( R ) ] db (Idesm) d [ID y(R)I db (IdesmSz)
MDIDD 21 (vw) 22 8 (w) 17 17
TDI/EG 14.5 12.5 12 12
a w, weak maximum; vw,very weak maximum.
Braggs law.
SMALL-ANGLE X-RAY SCATTERING STUDIES 1455

0.4

0.2

0 .o

-a2

r - (nm)

Fig. 11. Three-dimensional experimental correlation functions; MDIBD (open circles), TDI/EG
(open triangles).

a lamellar model (Fig. 12) show better-defined maxima for both samples. This
analysis yields a spacing of 12 nm for the TDI/EG material and 17 nm for the
MDI/BD material. These values are in excellent agreement with those deter-
mined by Bragg’s law, as shown in Table 11.

0 5 10 15 20 25
r (nm)
Fig. 12. Experimental correlation functions for one-dimensional lamellar model; MDIBD (open
circles), TDI/EG (open triangles).
1456 KOBERSTEIN AND STEIN

Fig. 13. Two-dimensional model of lamellar clusters in crystallized polyethylene copolymers.


Reprinted from ref. 24, p. 469, by courtesty of Marcel Dekker, Inc.

Close inspection of the one-dimensional correlation function for MDIBD


suggests that there may be a bimodal distribution of interdomain spacings, as
a weak maximum is seen at 8 nm. The breadth of the maximum in the scattering
profile for this material (Fig. 8) also suggests a distribution of lamellar spacings.
A distribution of spacings is reasonable in light of the DSC behavior, which
pointed to the existence of regions of varying degree of hard-segment ordering.
If such regions are localized, the domain structure would not be homogeneous,
and would provide a distribution of interdomain spacings. The model of Wilkes
and Yuseklg for oriented polyurethanes is similar in this respect.
A distribution of domain thicknesses would also be expected with the occur-
rence of segregation or fractionation of hard segments according to their sequence
length. There is evidence for the presence of a distribution in crystal thicknesses
in partially crystalline copolymers of polyethylene. SAXS studies of these
materials24led to the polycrystalline model shown in Figure 13. These copoly-
mers are similar to crystallizable polyurethanes in that the crystallizable se-
quences are of varying length and are located at varying intervals along the
SMALL-ANGLE X-RAY SCATTERING STUDIES 1457

molecule. Structural inhomogeneity of this type may be enhanced in the case


of bulk polymerization of polyurethanes, where phase separation during poly-
merization is known to O C C U ~ . Recent
~ ~ studies of this phenomenon have dem-
onstrated that heterogeneity in chemical composition of copolymer molecules
is found.26 Analysis of sol and gel fractions of bulk-polymerized polyurethanes
revealed a bimodal distribution of hard-segment content of the copolymer chains.
The observation of appreciable light-scattering intensity for MDI/BD with
parallel polaroids and the absence of intensity for crossed p o l a r o i d ~indicates
~~
that there are long-range heterogeneities in this material. The exact structure
of these heterogeneities, however, are not known at this time.
The results for the TDI/EG material are much different. Both the maxima
in the scattering pattern (Fig. 8) and in the correlation function (Fig. 12) are very
sharp, indicating a narrow distribution in interdomain spacing. Furthermore,
the oscillations in the correlation function damp out slowly, indicating a high
degree of spatial coherence of the domains. DSC measurements on this material,
however, did not provide evidence for hard-segment ordering. Hydrogen
bonding studies on similar materials also provide clear evidence for a greater
degree of intersegmental ordering in MDI/BD polyurethanes2g30compared to
asymmetric TDI-based polyurethanes.
Increased hard-segment ordering is reflected in the values of the predominant
interdomain spacings as well. Although there are slight differences in soft seg-
ments, the primary difference of the two polyurethanes is in the diisocyanate
symmetry. The asymmetrical TDI/EG hard segment is not well ordered and
probably exists in a conformation approximated by a random coil. The MDIDD
hard segments on the other hand are partially ordered into a paracrystalline
structure. As a result of the ordering, the MDI/BD domain is thicker, and the
spacing between domains must be larger. The experimental results support this
conclusion.

DiffusePhase Boundary Thicknesses


A detailed investigation of diffuse phase boundary thicknesses was undertaken
using slit-smeared, desmeared, and pinhole scattering data. The specifics of
this analysis will be the subject of a future paper, and only a summary of the re-
sults is presented in Table 111.
The first value in the columns refers to the results of analysis on smeared data
using eqs. (26) and (27). The background intensity for the smeared analysis is
estimated from a plot of 7(s)s3vs. s3 as previously discussed. The thickness of
the linear gradient is not obtained from an independent experiment ( E = f i a )
but serves as a more physically understandable value. Experimental plots de-

TABLE 111
Diffuse Phase Boundarv Thicknesses
Sample Sigmoidal gradient u (nm) Linear gradient, E (nm)*
MDI/BD 1.0 (1.1)(1.1) 3.3 (3.67) (3.67)
TDI/EG 0.6 (0.6) 2.1 (1.9)
* E = m o .
1458 KOBERSTEIN AND STEIN

v
rn
v)
m
-6.9 .
I \

-7.3 I I I I I

0 .01 .02 .03 .04

1.81 ( nm-l.81 1
Fig. 14. Experimental plot for determination of the diffuse phase boundary width for MDI/
BD.

picting the smeared analysis appear in Figures 14 and 15. The downward cur-
vature of the plots at small s is an indication that the Porod’s-law limit has not
yet been reached. The linear region that follows is where the asymptotic laws
hold. In the case of the TDI/EG polyurethane two linear regions are apparent
and the correct one must be chosen. Ruland31 has given a criterion for deter-
mining the minimum s value a t which Porod’s law holds. This criterion shows
that the analysis should hold at s values larger than that corresponding roughly
to the position of a secondary maximum (Imax-2 for this sample). In addition,
consideration of the slit optics provides a maximum value s,, above which the
infinite-slit assumption does not hold.32 Examination of these two limits (see
Fig. 15) clearly shows that the second linear region is the correct one for the
analysis.
Diffuse boundary thicknesses determined by analysis of desmeared and pin-
hole data are also presented in Table 111. Values obtained from desmeared data
appear in the first set of parentheses, while values from pinhole analysis are found
in the second set of parentheses. Background intensities were calculated from
plots of I(s)s4vs. s4 and diffuse boundary thicknesses were obtained through
application of (12) and (14). Values obtained through the various analyses are
in excellent agreement. The diffuse boundary zone for the MDI/BD polyure-
thane (linear gradient model) is ca. 3.3 nm wide compared to 2.1 nm for the
TDI/EG material.
The interpretation of diffuse boundary thicknesses in polyurethanes is not
a straightforward procedure. There exist a number of morphological features
that could contribute to an “effective” diffuse boundary thickness in addition
to the existence of a true concentration gradient. It is expected, for instance,
that the surface of polyurethane hard-segment domains is irregular owing to the
sequence length distribution. These features are all known to affect the observed
diffuse boundary thickness.33 In addition, the domains are expected to be ir-
regular in thickness and may have some curvature. Finally, if the distribution
SMALL-ANGLE X-RAY SCATTERING STUDIES 1459

si.ei (nm-i.ei)

Fig. 15. Experimental plot for determination of the diffuse phase boundary width for TDI/
EG.

of lamellar thicknesses is broad and their size is comparable to the diffuse


boundary thickness, the modeling used begins to break down. The results of
our measurements indicate that the analysis procedure may be near this point
of breakdown in the validity of the method since values of (r obtained are ap-
proaching the thickness of the domain. The agreement among the varying
analyses, and the excellence of fits obtained with the experimental data, however,
do provide a certain degree of confidence in the validity of the modeling. At least,
it is believed that the relative magnitudes of the values for MDIDD and TDIEG
are correct, that for MDI/BD being larger.

Electron Density Variance Calculations


Electron density variances were determined for the electron density profiles
depicted in Figure 3. The analysis was performed on both the Kratky slit-
smeared and the desmeared data following (17), (18),(28), and (29). In addition
the variance 5:for the case of complete phase separation was calculated from
(16). The values of volume fractions and electron densities needed for this
calculation were determined from the chemical composition and mass density
measurements.
Table IV contains the results of the electron variance calculations. Excellent
agreement is found between the results from the smeared and desmeared anal-
yses. In all cases the experimental variances q2’ (corrected for thermal density
fluctuations) and G2” (corrected for thermal density fluctuations and diffuse
boundaries) are smaller than the calculated variance G:. This indicates that
an appreciable amount of phase mixing occurs. The variance information can
be placed on a semiquantitative basis by comparing variances following the
method of Bonart.loJ1 These ratios are found in the last three columns of Table
IV. The degree of overall phase separation is 0.40 for sample MDIDD and 0.30
for sample TDI/EG. The maximum value for this ratio is 1.0 for the case of
complete phase separation. Both materials are therefore far from a state of
complete segregation, but the MDIBD sample possesses a greater degree of
phase separation than the TDI/E!G sample.
TABLE IV
x
0
Electron Density Variances* m
M
0v er a 11 Diffuse
phase boundary
E.3
- - - separation
- -mixing Domain
- -mixing 5
Sample Ap2" AP: Ap2' Ap"'lGZ Ap2"/G2'- 1 ApZ/Ap2"- 1 z
MDI/BD (smeared) 4.05 E-03 5.58 E-03 2.22 E-03 0.40 0.83 0.38
MDI/BD (desmeared) 4.16 E-03 5.58 E-03 2.26 E-03 0.41 0.84 0.42 $
T D I B G (smeared) 4.52 E-03 11.38 E-03 3.50 E-03 0.31 0.29 1.52
SMALL-ANGLE X-RAY SCATTERING STUDIES 1461

It is interesting to compare the results for the amount of diffuse boundary


mixing given in column 6. The value of this function is zero in the absence of
diffuse phase boundaries and increases with mixing in the interphase. The
MDI/BD material contains a larger amount of this type of mixing even though
it has a higher overall degree of phase separation. This result is more easily
understood upon examination of the amounts of mixing within domains shown
in the last column of Table IV. This function has a value of zero if the interior
of the domain is pure and increases as the purity decreases owing to mixing.
Experimental values indicate that the domains are relatively pure in the MDI/BD
polyurethane, but appreciably phase mixed in the TDI/BD material. This
contribution of mixing is sufficientlylarge to make the TDIBG material the more
phase mixed of the two on an overall basis, even though the phase boundaries
are sharper.
The increase in phase mixing within domains for the TDI/EG material can
be understood on thermodynamic grounds by consideration of the aromaticity
of the diisocyanate residues in the hard segment. The TDI entity has only half
of the aromatic character (i.e., half as many aromatic rings) as that of the MDI
precursor. On these simple grounds, the TDI hard segment should be more
compatible with the soft-segment polyether. It is therefore likely that the short
TDIBG sequences are dissolved in the soft-segment matrix, resulting in a higher
degree of mixing within domains.

Electron Density Profiles


Based upon the results of the SAXS analyses, a semiquantitative “picture”
of the domains may be reconstructed in terms of normalized electron density
profiles. The results of the SAXS analyses provided estimates of the diffuse
boundary thickness, interdomain spacing, and domain purity. Coupled with
the knowledge of pure-component electron densities and sample compositions,
the electron density profile for the lamellar model is obtained by invoking the
equality

This simply states that the integral of the normalized electron density profile
over one repeat unit (of length d ) must equal the volume fraction @HS of hard
segment. This is similar to a SAXS calculation of the linear crystallinity. To
fix this equality, there is one adjustable parameter, the hard-segment domain
thickness tHS. That is, for a lamellar model with sigmoidal interface density
gradient, the electron density is given by
~ ( r=) (PHS - PssNl - erf[(r - ‘/2t~s)/aIl+PSS (31)
where PHS is the hard-segment electron density, pss is the soft-segment electron
density, and a is the diffuse boundary thickness parameter. The error function
is defined as

The electron density profiles are determined then by finding a value for tHS
1462 KOBERSTEIN AND STEIN

TDI/EG SYSTEM

MDVBD SYSTEM

Fig. 16. Reconstructed experimental electron density profiles for MDI/BD and TDI/EG.

such that eq. (30) is satisfied. The results of this procedure appear in Figure
16, where the normalized electron density has been plotted as a function of the
distance r perpendicular to the domains. It is important to realize that these
profiles are not exact owing to several assumptions that were necessary to their
reconstruction. It was assumed that domains are lamellar and that the center
of domains are pure. Electron density variance calculations show this latter
assumption to be true for MDIBD, but not for TDI/EG. Electron density values
have therefore been omitted from the ordinate axis. It must also be emphasized
that the profile obtained reflects only an average value integrated over the dis-
tribution of domain thicknesses. The profiles are useful semiquantitatively in
depicting the microdomain structure.

Dynamic Mechanical Measurements


The temperature dependence of the tensile loss and storage moduli (E” and
E’) for the two polyurethanes is depicted in Figure 17. In the region of tem-
perature observed, both samples show only a single relaxation identifiable as
the soft-segment a-loss process. The transition is accompanied by a decrease
of two orders of magnitude in the modulus. The a-loss process itself looks similar
for the two materials and is associated with the glass transition of the soft seg-
ments. Further inspection of the figure reveals that the plateau modulus, for
temperatures above the a transition temperature, remains higher for the MDI/
BD polyurethane with symmetric hard segment. This elevation in modulus can
be attributed to a greater degree of phase separation and improvement of
hard-segment ~ r g a n i z a t i o n ~
in~this
- ~ ~material. With less phase mixing, fewer
hard segments are dissolved in the soft segments, resulting in a higher effective
physical crosslink density and thus an increase in modulus. In addition, seg-
mental ordering leads to a thicker domain and thus a higher number of chains
per unit interface, This effect also leads to an increase in modulus.
Examination of the tan6 peaks (Fig. 18) show that the intensity is of lower
magnitude for the MDI/BD material. This can also be attributed to improve-
ment in hard-segment organization which, would lead to restrictions on the
soft-segment motion^.^^-^^ If this restriction mechanism were the only factor
affecting the a-loss process, this decrease in intensity would be accompanied
by an increase in the peak temperature and a broadening of the peak. Results
SMALL-ANGLE X-RAY SCATTERING STUDIES 1463

0
0 0
I I I 1 1 I
0

-10000 -7500 -50.00 25.00 0 25.00 50.00 75.00

Temperature ("C)
Fig. 17. Temperature dependence of the storage and loss modulus for MDIBD (open circles)
and TDI/EG (open triangles).

indicate, however, that this is not the case. The peak temperature is lower for
the MDI/BD polyurethane. This is easily understood in terms of the SAXS
results, indicating an increased amount of phase mixing for the TDI/EG poly-
urethane. This increased mixing provides for an antiplasticization of the soft
segment. Two additional effects that may have some influence as well on this
temperature are the differences in molecular weight and degree of crosslinking
of the two soft segments, and the observed differences in diffuse boundary
thickness. It is therefore difficult to quantify the origins of this peak shift. The
observed result, however, is consistent with the increase in phase mixing found
by SAXS.

06
TAN ov A
O A
O A

TEMPERATURE - ("Q
Fig. 18. Temperature dependence of the loss tangent for MDI/BD (open circles) and TDI/EG
(open triangles).
1464 KOBERSTEIN AND STEIN

R EFLECT I NG
LATTICE PLANES

\
- 9.2A

-9.2;

Fig. 19. Bonart model for morphology of an elongated elastomer. Reprinted from ref. 17, p. 352,
by courtesy of Marcel Dekker, Inc.
SMALL-ANGLE X-RAY SCATTERING STUDIES 1465

I I
10-

08 -

06 -
NX

04-

02 -

O' 2 4 6 8 (0 1'2 1'4 16 :8

x - Diisocyanate Residues per


Hard Segment
Fig. 20. Theoretical hard-segment length distribution function.

A Model for Microdomain Structure


A number of models for polyurethane microdomain structure have appeared
in the literature. 10,11,19,37-39
These models all have many common features with
the original model of Bonartl7 shown in Figure 19. Hard-segment sequences
are pictured as rigid rods with preferred orientations as a result of strong hy-
drogen bond interactions. The Bonart models, however, were originally con-
ceived in order to explain the wide-angle diffraction peaks observed in oriented
heat-set polyurethanes. A different morphology is to be expected in bulk elas-
tomers. Furthermore, these models neglected an important feature of segmented
polyurethane elastomers, the hard-segment sequence length distribution. This
is a major consideration to include in an improved microdomain model, and may
in fact be a dominating factor.
The hard-segment sequence length distribution for polyurethane block co-
polymers has been shown to be a most probable distribution.*O The calculated
distribution functions for MDI/BD and TDI/EG (assuming 100%conversion)
appear in Figure 20. The important feature to note is that a large fraction of
the sequences are very short. Since electron variance calculations for MDI/BD
showed that there is little phase mixing within domains, these short sequences
must be incorporated within the hard-segment domains.
Assuming that hard segments are rigid rods, as has been proposed in previous
models, a hard-segment packing model based on this distribution function can
be constructed. Figure 21 shows such a two-dimensionalpacking model, in which
the hard segments were placed in a one-dimensional array by a random number
generator which was weighted by the sequence length distribution function
truncated a t a value of 18units per sequence. In the same figure, the normalized
electron density profile associated with this spatial distribution of hard segments
is presented. The theoretical electron density profile is in direct conflict with
the experimental SAXS results (Fig. 16) since this distribution function produces
an extremely large diffuse boundary thickness. In fact, the diffuse boundary
is larger than the domain spacing.
It is evident that some modifications to the Bonart model are necessary to
1466 KOBERSTEIN AND STEIN

Hard Segment Packing Model

-160 -120 -00 -40 0 40


. 00 120 16(
Distance-r(n)

Fig. 21. Theoretical electron density model for random rodlike hard-segment packing model.

explain the results of this study. The first of these is consideration of the se-
quence length distribution as discussed above. In addition, it is apparent from
comparing Figures 16 and 21 that the extended-chain conformations inherent
to previous models do not conform to the experimental results. The ex-
tended-chain conformations assumed in previous models are not realistic on
entropic grounds, and suffer from the same problems as the fringed micelle
model. A more realistic explanation for the observed electron density profiles
is that chains turn back into the domains. In this fashion, chains may order
preferentially as a result of hydrogen bonding, yet still accommodate short se-
quences into the domains.
The ability of chains to turn back into the domains depends on the chain
flexibility or in other words the persistence length of the hard-segment sequence.
This information is not available, but an idea of flexibility may be obtained by
examining structural schematics of the two hard-segment repeat units as shown
in Figure 22. The chain extenders (BD and EG) are relatively flexible and are
depicted as springs on either side of a central chevron. The diisocyanate residues
on the other hand are relatively rigid. The urethane group is planar, and thus
the 4 and x rotations result in only small changes of size for this unit. The size
of this rigid unit is appreciably larger for MDI than for TDI, and the MDI unit
is symmetrical and capable of regular packing, while the TDI unit is not. The
flexible chain extender is also longer for the MDI copolymer, which can com-
pensate for the rigidity of the diisocyanate residue. These simple conformational
considerations show that coiled-chain conformations are possible and that ex-
tended-chain conformations are not necessarily favored.
SMALL-ANGLE X-RAY SCATTERING STUDIES 1467

MDI/ BD
H\ /H

Fig. 22. Schematic representationsof the chemical structure and rigidity of MDIDD and TDI/EG
hard-segment repeat units.

If hard-segment chains do turn back on themselves, the exact manner in which


this takes place will exert a strong influence on the morphology. The type of
chain coiling in turn would be highly dependent on the segment-segment in-
teractions inherent to the hard segment. If there are only weak interactions the
chains will have conformations approaching Gaussian. If strong interactions
are present, more extended conformations can be assumed as the hard segments
order preferentially. Finally, if the sequence is crystallizable, a chain-folded
lamellar crystal may be obtained. These arguments can be applied directly to
MDI/BD and TDI/EG to help explain the morphological differences resulting
from the change in diisocyanate symmetry. In the case of the symmetric
MDI/BD the segment-segment interactions are much stronger and more ex-
tended-chain conformations are expected to be favored. This is reflected directly
in the larger values of interdomain spacings for MDI/BD.
Construction of a more quantitative microdomain model requires assumption
of a model for the domain geometry. For these purposes the lamellar poly-
crystalline microdomain morphology represented in Figure 13 has been adopted.
This is an approximation of the real morphology but appears to be the best ap-
proximation. A random two-phase morpyhology may be ruled out because of
the maximum observed in the scattering curve. The great attraction of the la-
mellar model is that it allows for simple incorporation of the short sequences as
well as the long sequences into the same domain. It also provides a smooth
morphological tansition to the lamellar spherulitic morphology observed in
semicrystalline polyurethanes. MDI/BD polyurethanes, for example, are known
1468 KOBERSTEIN AND STEIN

fq MDI/BD
~~~~~~
TDl/EG
,

-4 -2 0 2 4 -4 -2 0 2 4

DOMAIN

Fig. 23. Two-dimensional schematic intradomain segmental ordering models and corresponding
experimental electron density profiles for MDI/BD and T D I E G .

to crystallize upon annealing23and to exhibit lamellar crystalline spherulitic


s u p e r s t r ~ c t u r e .In
~ ~polyurethanes, owing to the complexity of their chemical
structure and interactions, various departures from the lamellar model are ex-
pected to occur. Hard-segment domains are probably more like ribbons that
have finite width, undulating thickness, rough surfaces, and various twists,
curvatures, and tilts as they traverse the material. Treatments of lattice dis-
tortions of these types have a p ~ e a r e dbut ~ ~are
, ~difficult
~ to apply in practice.
In view of these difficulties, the lamellar model is applied directly to these sys-
tems, with the reservation that it is an approximation to the real morphology.
The lamellar domain model does conform well to the experimentally deter-
mined electron density profile for MDI/BD shown in Figure 16. In this figure
the hard segments are seen to have a thickness of approximately 4 nm. This
value is corroborated by direct analysis of the one-dimensional correlation
function by the method of Strob1,42which also yields an average domain thickness
of 4 nm. This distance corresponds well to the “stem length” of a chain consisting
of about two MDI units traversing roughly perpendicular to the domain surface.
Two MDI units is also the most probable sequence length, as shown in Figure
20. The short sequences are thus the dominant factor that influences domain
morphology in that they impose a lamellar geometry upon the system, and
control, to a large extent, the domain thickness.
Models taking into account the amount of intradomain hard-segment order
can be constructed which demonstrate qualitatively the features of the experi-
mental electron density profiles of Figure 16. These models, based on the
schematic representations of hard-segment structure (Fig. 22), appear in Figure
23.
SMALL-ANGLE X-RAY SCATTERING STUDIES 1469

In the model, hard segments within the MDI/BD domain are aligned prefer-
entially as a result of their symmetry. In this “average” view of a domain, the
longer hard segments are forced to fold back into the domain. The presence of
segregation might be expected to decrease the amount of chain reentry into the
domain; however, this effect might also lead to the formation of folded-chain
crystals.
It is not unreasonable to assume that the MDI/BD hard-segment sequences
behave in a fashion similar to other crystallizable polymers. The morphology
of crystalline urethane copolymers would then be similar to that of other block
copolymers with a crystallizable sequence such as poly(styrene-b-ethyleneoxide).
Morphological studies of such systems43have shown that the morphology is in-
deed lamellar. Certain polyamides, similar to polyurethanes in their chemical
structure and hydrogen bonding behavior, are also known to chain fold.44 The
key difference between such systems and the polyurethanes studied here is the
sequence length distribution. In this case the lamellar thickness is to a large
extent limited by accommodation of the short sequences into the domain.
Strong support for chain folding of ordered hard-segment sequences was
presented in a recent investigation of hard-segment model compounds.21
Measurements of melting temperatures of MDI/BD hard-segment analogs were
taken as a function of sequence length. These measurements showed a maxi-
mum melting temperature at three MDI units per sequence. The decrease in
T , above three units was attributed to the incorporation of defects into the
crystals due to the occurrence of chain folding. It is interesting to note that three
MDI units corresponds well to the thickness of a hard-segment domain as de-
termined from the SAXS measurements.
It may even be possible for hard segments to chain fold as tight adjacent folds
in the crystalline lamellae through introduction of gauche conformations into
the butanediol residue. Assuming that the unit cell of B l a ~ k w e lmay l ~ ~ be ap-
plied, the dimensions of the a and b axes would be 5.05 and 4.67 A,respectively,
for the MDI/BD crystal lattice. The introduction of a single gauche bond into
the O-(CH2)4-0 unit results in a length of 5.34 A,which could conceivably
be accommodated as a tight chain fold.
The model presented in Figure 23 may also provide a reason for the absence
of WAXS diffraction in MDI/BD. The actual conformation of the hard-segment
chains is visualized (Fig. 24) as being similar to that of crystalline MDI/BD,46
with the exception that gauche sequences are included in the butanediol residue.
These gauche sequences lead to chains reentering the hard-segment domain with
incorporation of many defects that lead to poorly defined diffraction maxima.
In addition, accommodation of the short sequences in the domain leads to thin
domains, essentially restricting the crystal lattice to two dimensions. In this
respect the chains can be considered to possess a more liquid-crystalline or
perhaps paracrystalline order, and only weak wide-angle diffraction would be
observed even though the system would show endothermic DSC behavior as the
order melts out. This corresponds well to the behavior of the MDI/BD
sample.
Analysis of the various measurements on the TDI/EG system lead to a quite
different morphology represented schematically in Figure 23. In this case DSC
measurements do not support the existence of appreciable hard-segment ordering
as no endothermic activity was observed. The asymmetry of the TDI residue
1470 KOBERSTEIN AND STEIN

Fig. 24. ac projection of poly(MD1-butanediol) hard segments as proposed by Blackwell and


Gardner. Reprinted from ref. 46, p. 16, by courtesy of IPC Business Press.

inhibits ordering and subsequently does not produce significant chain alignment
perpendicular to the domain. The chains are less restricted in their conforma-
tions and do not protrude from the domains to the same extent. The confor-
mation of the T D I E G sequence is thus closer to that of an amorphous block
copolymer sequence although some order exists, owing to the specificity of hy-
drogen bond interactions. The electron density profile in Figure 15, however,
again shows that the thickness of the domain is small enough to accommodate
some of the short hard-segment sequences. In contrast to the MDIDD sample,
electron density calculations revealed that domains were phase mixed. The
phase mixing occurs as a result of the higher relative compatibility in TDI/EG
compared to MDI/BD. This is a result of the decreased aromatic character of
TDI compared to MDI, as discussed above, and probably leads to the dissolution
of some of the short hard-segment sequences in the soft-segment domain.
The calculated domain thickness for T D I E G supports this interpretation.
In contrast to MDIDD, the domain thickness of ca. 3 nm is equivalent to about
three TDI units. This value implies that the hard segments containing two TDI
units are dissolved in the soft segment, and thus the next longest hard-segment
sequence controls the domain thickness. If this reasoning is correct, the effect
SMALL-ANGLE X-RAY SCATTERING STUDIES 1471

of compatibility on microdomain structure is clear: the shortest nonsoluble hard


segment controls the domain thickness. Additional experiments are needed,
however, to verify this hypothesis.
The absence of strong preferential .ordering of TDI/EG hard segments has a
strong influence on the domain morphology. The interdomain spacing is smaller
than in MDI/BD. This is an expected result from a lamellar model since more
ordering necessarily leads to more extended conformations and hence a larger
interdomain spacing. The differences in correlation functions between the two
samples may be understood from the same arguments. In the absence of crys-
tallinity (i.e., in TDI/EG), a distribution of crystal thickness is not realized and
the spatial coherence of the domains is more long range. The resultant scattering
maximum is therefore significantly narrower (Fig. 8) and the maxima in the
correlation function are well defined and damp out more slowly (Fig. 12).
Finally, these simple structural schematics clearly explain the broader diffuse
boundary zone found in MDI/BD. The broader phase boundary results from
two factors. The first factor is the longer length of the statistical segment length
in the MDIDD sequence, which for an amorphous block copolymer would lead
to a larger diffuse boundary width according to present the0ries.4~ The second
factor is that the strong interurethane bonding and ordering in the symmetrical
MDI/BD sequence, coupled with the bulky nature of the units, leads to protrusion
of some units from the domain as they fold or coil back into the hard-segment
domain. Both of these factors contribute to a larger diffuse boundary thickness
for the MDI/BD polyurethane.
In summary, the schematic models presented for microdomain structure do
account qualitatively for all of the experimental observations pertaining to the
two polyurethanes studies in this investigation. The models are not intended
to be exact nor are they expected to apply universally for all polyurethanes. They
do, however, provide plausible explanations for the effects of diisocyanate
symmetry, diisocyanate length, and hard-soft segment compatibility on hard-
segment microdomain structure. The value of thse models is that they identify
the future experiments necessary to verify the interpretations presented here.
It is clear that these experiments must involve carefully chosen and prepared
model polyurethane systems in order to obtain a clear and quantitative description
of the factors that influence polyurethane microdomain structure.
Supported by grants from the General Tire and Rubber Company, the National Science Foun-
dation, and the Materials Laboratory of the University of Massachusetts. One of the authors (J.T.K.)
wishes to acknowledge partial support of this work under a National Science Foundation grant
(DMR-8105612).

References
1. G. A. Senich, Ph.D. Thesis, University of Massachusetts, 1978.
2. R. W. Hendricks, J . Appl. Crystallogr., 11,15 (1978).
3. E. J. Roche, Ph.D. Thesis, University of Massachusetts, 1978.
4. P. Debye and A. M. Bueche, J . Appl. Phys., 20,518 (1949).
5. A. Guinier and G. Fournet, Small-Angle Scattering of X Rays, Wiley, New York, 1955.
6 . G. Porod, Kolloid Z., 124,83 (1951); 125,51 (1952); 125,108 (1952).
7. W. Ruland, J.Appl. Crystallogr., 4,70 (1971).
8. C. G. Vonk, J. Appl. Crystallogr., 6,81 (1973).
9. W. Wiegand and W. Ruland, Prog. Colloid Polym. Sci., 66,355 (1979).
1472 KOBERSTEIN AND STEIN

10. R. Bonart and E. H. Muller, J . Macromol. Sci. Phys., BlO, 177 (1974).
11. R. Bonart and E. H. Muller, J . Macromol. Sci. Phys., BlO, 345 (1974).
12. J. T. Koberstein, B. Morra, and R. S. Stein, J. Appl. Crystallogr., 13,34 (1980).
13. N. S. Schneider, C. S. Paik Sung, R. W. Matton, and J. L. Illinger, Macromolecules, 8, 62
(1975).
14. S. B. Clough and N. S. Schneider, J. Macromol. Sci. Phys., B2,553 (1968).
15. R. W. Seymour and S. L. Cooper, Macromolecules, 6,48 (1973).
16. G. W. Miller and J. H. Suanders, J. Polym. Sci. A-1,8,1923 (1970).
17. R. Bonart, L. Morbitzer, and G. Hentze, J. Macromol. Sci. Phys. B3,337 (1969).
18. R. Bonart, L. Morbitzer, and E. H. Muller, J. Macromol. Sci. Phys., B3,447 (1974).
19. C. E. Wilkes and C. S. Yusek, J . Macromol. Sci.Phys., B7,157 (1973).
20. W. J. MacKnight, M. Yang, and T. Kajiyama, Polym. Prepr. A m Chem. SOC.Diu. Polym.
Chem., 9,860 (1968).
21. Y. Camberlin, J. P. Pascault, M. Letoff6, and P. Claudy, J . Polym. Sci. Polym. Chem. Ed.,
20,383 (1982).
22. G. W. Miller, J . Appl. Polym. Sci., 15,39 (1971).
23. D. S. Huh, Ph.D. Thesis, University of Wisconsin, Microfilm #71-28,992.
24. H. G. Kilian and W. Wenig, J . Macromol. Sci. Phys., B9,463 (1974).
25. M. Tirrell, L. J. Lee, and C. Macosko, Am. Chem. SOC.Symp. Ser. 104,149 (1979).
26. W. J. MacKnight, unpublished.
27. J. T. Koberstein and R. S. Stein, to appear.
28. R. W. Seymour, G. M. Estes, and S. L. Cooper, Macromolecules, 3,579 (1970).
29. C. S. P. Sung and N. S. Schneider, Macromolecules, 8,68 (1975).
30. C. M. Brunette, S. L. Hsu, and W. J. MacKnight, Polym. Prepr. Am. Chem. SOC.Diu. Polym.
Chem., 22,353 (1981).
31. W. Ruland, Colloid Polym. Sci., 255,417 (1977).
32. L. B. Shaffer and R. W. Hendricks, ORNL-TM-4278, Oak Ridge National Laboratory,
1973.
33. C. G. Vonk, J. Appl. Crystallogr. 11,541 (1978).
34. D. S. Huh and S. L. Cooper, Polym. Eng. Sci., 11,369 (1971).
35. S. L. Cooper, J. C. West, and R. W. Seymour, Ency. Polym. s'ci. Technol., Suppl. 1, 521
(1976).
36. C. M. Brunette, S. L. Hsu, W. J. MacKnight, and N. S. Schneider, Polym. Eng. Sci., 21,163
(1981).
37. S. L. Samuels and G. L. Wilkes, J. Polym. Sci. Polym. Symp., 43,149 (1973).
38. N. S. Schneider, C. R. Desper, J. L. Illinger, A. 0. King, and D. Barr, J. Macromol. Sci. Phys.,
B11,527 (1975).
39. Z. Ophir and G. L. Wilkes, J. Polym. Sci. Polym. Phys. Ed., 18,1469 (1980).
40. L. H. Peebles, Jr., Macromolecules, 7,872 (1974); 9,58 (1976).
41. I. D. Friedman, E. L. Thomas, L. J. Lee, and C. W. Macosko, Polymer, 21,388 (1980).
42. G. R. Strobl and M. Schneider, J . Polym. Sci. Polym. Phys. Ed., 18,1343 (1980).
43. B. Lotz, A. J. Kovacs, G. A. Bassett, and A. Keller, Kolloid 2.2.Polym., 209,115 (1966).
44. P. F. Dismore and W. 0. Statton, J. Polym. Sci. Part C, 13,133 (1966).
45. J. Blackwell and M. R. Nagarajan, Polymer, 22,202 (1981).
46. J. Blackwell and K. H. Gardner, Polymer, 20,13 (1979).
47. E. Helfand, Acc. Chem. Res., 8,295 (1975).

Received August 23,1982


Accepted February 18,1983

You might also like