Vous êtes sur la page 1sur 31

JOURNAL OF POLYMER SCIENCE: PART A-2 VOL.

10,1415-1445 (1972)

Fracture Behavior in Nylon 6 Fibers

B. A. LLOYD, K. L. DEVRIES; and M. L. WILLIAMSl College of


Engineering, University of Utah, Salt Lake City, Utah 84112

synopsis
A reaction rate model of fracture in polymer fibers is described. This model assumes
that bond rupture is governed by absolute reaction rate theory with a stress-aided
activation energy. It is demonstrated that the key in obtaining good agreement be-
tween the model and experiment lies in taking proper account of the variation of stress
on the tie-chain molecules. The more taut chains rupture first, and the load is redis-
tributed among the remaining unruptured tie chains. The effect of varying the tem-
perature both in the model and in experiments on fracture in fibers is explored. Good
agreement between predictions of the model and experiment is possible only with an un-
deterstanding of the distribution in stress on the tie chains. The distribution in stress on
the chains was experimentally determined by monitoring the kinetics of bond rupture
with electron paramagnetic resonance (EPR) spectroscopy. Temperature is found to
have two effects on macroscopic strength. (1) The thermal energy aids the atomic
stress in breaking the atomic bonds; as a consequence the rate of bond rupture of a family
of bonds under a given molecular stress is increased. In this respect temperature might
be viewed as decreasing the “strength” of a bond. (2) Temperature also serves to
“loosen” the molecular structure and in this way modify the distribution in stress on the
tie chains. To explain bond rupture and macroscopic fracture behavior quantitatively,
account must be taken of both effects.

INTRODUCTION
This study is limited to a very narrow and yet important class of polymers,
i.e., the highly oriented semicrystalline polymer fibers. Several models
have been proposed to describe the microstructure in highly oriented fibrous
polymers. These range from the early fringed micelle structures’ to more
recently proposed folded chain model~.~JAlthough models proposed by
Hearlej4Bosley15Keller? Fischer et al.,? Peterlin12Dismore and Statton:
Gubanov and ChevychelovlgTakayanagi et al.,1° and Bonart and Hose-
mann3 differ in some very important details, they have a basic common
element in their microstructure; that is, all include a “sandwich” type
structure alternativg between highly ordered regions and less ordered or
“tie molecule” regions. For the particular melt spun, hot-drawn nylon 6
fibers used in this research, Joon Park’l has recently proposed a para-
crystalline, folded chain structure as representative of the fiber micro-
structure.
Many researchers have proposed theories and molecular models to
explain fracture from a microscopic viewpoint.12-*5 However, theoretical
1415
@ 1972 by John Wdey & Sons, Inc.
1416 LLOYD, DEVRIES, AND WILLIAMS

predictions of failure on a microscopic scale could not be experimentally


verified until 1964, when Zhurkov and his associates demonstrated that
electron paramagnetic resonance (EPR) could be used to measure chain
scission caused by mechanical stress in polymers.16 Since that time, the
authors and a number of other researchers have extended these studies
with the goal of establishing a satisfactory failure model for fibers.
I n retrospect, it appears that many past molecular models of fracture had
a common oversimplification. They neglected the difference in stress
among the polymer chains. To successfully correlate bond rupture kinetics,
one must assume the more highly-stressed bonds rupture first and the load
must be redistributed among the remaining chains.
Reaction rate theory is well established and any model of failure should
include its features. The probability Pb of bond rupture would then be
given by
P b = woexpt - (u,- y u ) / k ~ } (1)
where wo, Uo,and y are kinetic parameters commonly known as the collision
parameter, the activation energy, and activation volume, respectively. The
key to successful use of such theory appears to lie in the use of the stress
term u and its relationship to the fiber microstructure. Inspection of eq.
(1) reveals that u is the atomic stress on a bond in a polymer chain and it
appears likely that this stress is not equal (or proportional independent of
time) to the applied macroscopic stress. I n fact, as it is most unlikely that
all chains in the polymer are equally stressed, there would exist a distribu-
tion of stresses related in some fashion to a distribution of “effective” chain
lengths in the fiber microstructure. This major modification to the earlier
models was suggested by Roylance et al.,” Kausch et al.,l* Peterlin12and
Verma and Peterlin. l9 Further development of the statistical mathematical
approach to the problem was presented by Kausch.m,21
A more extensive study has been conducted by the authors and pre-
liminary results on nylon 6 fibers published.22 The preliminary results of
this study indicated a strong dependence of the fiber strength on the ap-
parent distribution of the load-carrying chains as observed by electron
paramagnetic resonance (EPR) techniques. Furthermore, it was observed
that this apparent distribution varied with temperature and material.
Permanent as well as reversible changes in the distribution could be made,
depending on the fiber treatment and temperature. This paper extends
the experiments and application of the model to other loadings and tem-
peratures. It discusses the effect of the distribution in stress among the
load-carrying chains and its dependence on temperature, treatment, and
materials. It presents in detail the extension of a reaction-rate theory
model to other temperatures and discusses the insight gained toward the
fracture process by comparison of theoretical and experimental data.
TEST FACILITIES AND PROCEDURES
The experimental equipment, sample preparations, and experimental
procedures were as previously reported.22 The major portions of this work
FRACTURE BEHAVIOR IN NYLON 6 1417

on oriented fibers has been conducted on nylon 6 having a number-average


molecular weight between 25,000 and 30,000 supplied by Allied Chemical
Corporation. These fibers, having a singlefilament diameter of approxi-
mately 0.001 in. (2.54 x cm) were wound in bundles having a cross-
sectional area of (15-16) x cm2. The Varian V-4257 variable tem-
perature accessory was used in conjunction with the Varian E-3 spectrome-
ter and the accessory was modified to allow the fiber sample to pass com-
pletely through the spectrometer cavity. The sample ends were enclosed
with a rubber “hat” and a positive helium pressure was maintained to
exclude air during testing. In this way the sample could be exposed to any
temperature between - 150°C and +200”C while being maintained in an
inert atmosphere.

BASIC MODEL
An effort has been made by several researchers to determine in which part
of the microstructure chain scission occurs during mechanical loading.
Verma and Peterlinlg have carefully studied the EPR spectra hyperfine
structure of radicals produced by irradiation and mechanical stress in
oriented and amorphous polymers. They concluded that the majority of
the free radicals generated during mechanical loading were located in mo-
lecular chain sectionswith little or no orientation. From this they concluded
that the radicals were located in the quasi-amorphous layers sandwiched
between the crystalline blocks. Further evidence supporting the basic
assumption that during mechanical loading chain scission occurs almost
wholly in the less ordered or tie molecular regions was reported by Becht
and F i ~ c h e r . These
~ ~ investigators observed that the nylon radicals pro-
duced by stressing were all converted to the methacrylic radical during
swelling with methacrylic acid. On the other hand, radicals produced by
irradiation are fairly uniformly distributed throughout the sample, and the
number converted to the methacrylic radical was found to correspond ap-
proximately to the amorphous fraction. Since the acid is thought to swell
only the amorphous part of the sample, it was concluded that mechanically
induced radicals were produced largely in the looser parts of the polymer
structure. Similar conclusions have been reached by observation of the
rate at which nylon radicals convert to peroxy radicals in the presence of
oxygen. Once again, a marked difference is observed in the peroxy be-
havior of stress-induced and radiation-induced radicals. The latter oc-
curred at markedly slower rates. There is a possibility that the looser
structure about the free radicals is produced by the bond rupture and was
not initially present at these sites. To our knowledge this question has not
been completely resolved as yet.
An important point to make here is that, even though chain scission is
considered to occur almost wholly in the more disordered or tie molecule
regions of the polymer fibers, there still remains the question: Does failure
occur in every tie molecular region or are there specific critical tie molecule
or flaw regions distributed throughout the fiber in which failure predom-
I20

100

%-
x
80
v

n
5

2
ii:
60
v)
a

40

20

120

1
2 4 6 8 10 12 14 16 18 20
STRAN (%I
(b1
Fig. 1. Histograms of nylon 6 samples from stepstrain EPR data at room
temperature: ( a ) nylon 6, #l; ( b ) nylon 6, #2 annealed.
FRACTURE BEHAVIOR IN NYLON 6 1419

Fig. 2. Probability plot of nylon 6 histogram data at room temperature (cumulative


per cent of estimated total spins/cma): ( 1 ) nylon 6, #l;(9)nylon 6, #2 annealed.

inates? Observation of the number of free radicals generated during frac-


ture has substantiated the concept of a failure process occurring throughout
the fiber at many critical flaw and analysis of this number in t e r m
of the total chain scissions that could occur should all tie regions be active
seem to indicate only a fraction of the number possible are generated and/or
observed by EPR.% Further discussion of this point will be made later;
however, there exists some strong evidence from this study and other favor-
ing the concept of regions acting like rheological units or ‘(crystalline
blocks” interspaced with (‘criticalflaw” regions to explain the strength and
failure properties of the polymer fibers. These crystalline-block regions
may in fact contain some of the less-ordered or tie-chain regions and yet
act elastically as an unit. The concept of a critical flaw region in which
1420 LLOYD, DEVRIES, AND WILLIAMS

failure predominates, i.e., a region that would perhaps contain the fewest
and/or the most poorly distributed load-bearing tie molecules, is in basic
agreement with previous observations in this laboratory. 17e24.26,* I n fact,
one might logically expect failure to occur in flaw regions following a selec-
tive path of “least resistance” around the more densely packed blocks.
An important concept that earlier models neglected in establishing a
failure criterion was the existence of a distribution of stresses within the
polymer fiber related in some fashion to a distribution of effective stressed
chain lengths in the critical regions. This concept of critical regions con-
sisting in part of a set of tie chains, all of different length and number, was
recently proposed by several a ~ t h o r ~ ~and J ~is Jthe~ essential
- ~ ~ ~key~ ~to
this study. The first experimental verification of such a distribution was
reported by Kausch and Becht,18where bond scission data from EPR step-
strain tests was plotted in the form of a histogram giving the number of
chain scissions versus the strain increment (chain length).
Figure 1 shows the results of step-strain room temperature tests of two
different nylon 6 samples plotted as histograms. It should be noted that,
due to the fact that the sample fails rather catastrophically, the tailing side
of the distribution is missing. These experimental studies, therefore, give
no information on the distribution in this region. On analyzing these re-
sults in terms of a mathematical distribution function, it was necessary to
assume some type of form to the missing portion. It was decided from
analysis of the data, some of which gave considerable portions of the tailing
side, that a symmetrical distribution was the most probable type. With
this assumption the observable portion of the results was plotted on prob-
ability paper and this portion, at least, was found to comply with a normal
or Gaussian distribution criterion. Figure 2 shows the results of the histo-
grams shown in Figure 1 plotted on probability paper. The criterion for a
Gaussian distribution is that these plots be straight lines. From these re-
sults an “apparent or effective distribution” in length was found to exist.
This effective distribution was found to be temperature-dependent as well
as material-dependent. A discussion of the temperature dependence is
presented in a subsequent section of this paper. For nylon 6 #1 a t room
temperature the distribution one standard deviation value was observed
to be approximately lOyoof the distribution mean strain value.
As yet, the current study has not been able critically to distinguish be-
tween the various morphological models of fibers, i.e., modified fringed mi-
celle, folded chain, paracrystalline, etc. It does point out several features
that a “correct” model must contain. (1) There must be weak “flaw” re-
gions reasonably uniformly distributed throughout the material. The
most direct proof of this feature is the observation that in nylon 66 monofila-
ment fibers stressed very near their fracture point the free radicals are dis-
tributed throughout the sample length. The number of radicals produced
during fracture also supports this observation. (2) Crack arrestors (densely
packed regions?) must be distributed about the flaw regions. Otherwise,
one or a very few of the most critical flaws would proceed to sample failure.
FRACTURE BEHAVIOR IN NYLON 6 1421

Such behavior would once again prevent the observations of statement one
above. (3) There must be a distribution in either the effective length of
the chains in a given flaw region or in the size (criticality) of the flaw regions.
The proposed model apparently does not distinguish between these two
possibilities.
One other possibility is perhaps worthy of note. There is as yet no direct
evidence to support it, but one can envision mechanisms where some poly-
meric chains might be broken without the development of observable free
radicals. For example, a number of free radicals might interact and thereby
annihilate each other or the pair of free radicals associated with a broken
chain might act as a catalyst for rupture of neighboring chains. Electron
exchange between the radicals and the new scission points would preclude
the development of additional free radicals. If this reaction were to con-
tinue, a pair of free radicals could represent a number of broken bonds.
Careful studies of free-radical decay conducted both in this laboratory and
elsewhere have indicated that up to room temperature and above free radi-
cals in nylon are fairly stable. The above mechanism might, however, not
be expected to show up in such decay tests. The authors’ considered opin-
ion is that the free radicals produced does closely represent the number of
chains broken. However, even if such is not the case and there are more
bonds broken than free radicals formed, one would logically expect the free
radical concentration to be proportional to the total number of chains rup-
tured. As a consequence this number would not deter from the proposed
basic model but might modify some of the parameters.
It should be made clear, as noted under statement three above, the ex-
perimentally determined distribution could be either of two phenomena or a
combination of them. First, there is the tie-chain distribution in a given
flaw and secondly, the distribution in size or criticality of flaws. To help
in analyzing the experimental evidence the following idealized model is
proposed.
The mathematical model incorporates a basic alternating sandwich struc-
ture of crystalline blocks and critical flaw regions. These regions have no
predetermined size and are related to each other by ratio values only. A
distribution in effective length of chains in the critical flaw regions is in-
cluded and, hence, different chains are under different stresses. The dis-
tribution in stress among the tie chains changes as bond rupture occurs and
the load carried by these chains must be redistributed among the remaining
unbroken chains. The fracture of the chains is assumed to be controlled by
a Tobolsky-Eyring12type rate process or reaction-rate theory, i.e., a stress-
aided activation energy process. The model takes one such crystalline
block and flaw region as representative of the average of all critical regions
of the fiber. The ordered or ‘(crystallineblock” region is represented by a
number of aligned polymer chains, each with an average bond area Ab.
The disordered or (‘criticalflaw region” is assumed composed of sets of tie
chains. Each of these n sets has a different effective length and number of
chains according to a prescribed distribution. The original length of the
1422 LLOYD, DEVRIES, AND WILLIAMS

flaw region is taken as los, and some of the tie chains are initially considered
to be taut whereas all other sets may be kinked, disoriented, folded, etc.
and do not become taut until the region is sufficientlystrained to remove the
slack for a particular set. The derivation for these equations has been
previously outlinedzzwith the basic assumptions: (1) the shear mode of
failure, i.e., chain slippage or flow due to secondary bond failure in the aver-
age of the critical regions is negligible compared to backbone chain scission;
(2) chain scission in the “crystalline block” region is negligible; (3) Hooke’s
law is applicable on an atomic scale; (4)a chain assumes no load-carrying
ability until the strain in the flaw region is such that the chain will be taut
and after it is taut, it is assumed to act elastically; (5) chain scission is pre-
dicted by kinetic reaction rate theory. The governing equations for an
applied macroscopic stress UT as a function of time were found for chain rup-
ture to be:

where
CBi = CuOf - Cur
where CBTdenotes the total broken polymer chains in the average of the
critical regions of the fiber at time t , CB*is the number of broken polymer
chains in the ith set at time t , and C,O, is the original number of polymer
chains in the ith set. The number of unbroken polymer chains in the ith
set, Cut,was determined by numerical solution of the set of equations:
dCut/dt = - CurA exp {Bu,) For i = 1, n (3)
where n is the total number of chain sets in critical region,
A = wo exp {- Uo/kT)
B = r/kT

and ui is the stress on each bond in the ith set in the critical region, k is
Boltzmann’s constant, T is the absolute temperature, w o is the collision
parameter, Uo is the activation energy, and y is the activation volume
The stress ut on each bond in the critical region was related to applied
macroscopic stress UT by a force balance. The stress and strain in the
crystalline block regions were
UC = UT (4)

eC = (5)
where Eb is the modulus of elasticity of a single polymer chain. The total
macroscopic strain is:
eT = (GEL + %)/(I+ EL) (6)
FRACTURE BEHAVIOR IN NYLON 6 1423

where R L is the ratio of original “cry-blk” length to original critical flaw


length and tais the strain in the flaw region.
It is important t o note several things about the derivation. The key
equations only contain ratio values of the idealized geometry and the bond
area Ab cancels out in the derivation. A similar derivation for an applied
macroscopic strain has been detailed previously by Lloyd.29
The input parameters necessary for a theoretical solution using the
mathematical model of the polymer fiber are: S = the standard deviation
of the distribution of chains in the critical region; Rc = the ratio of the
number of polymer chains in the “cry-blk” region to number of tie chains
in the flaw region; RL = the ratio of the original length of the “cry-blk”
region to the original length of the flaw region; E b = the modulus of elas-
ticity of a single polymer chain; w o = the collision parameter; Uo = the
activation energy; y = the activation volume; UT or ET = the applied
macroscopic stress (or strain) as a function of time; T = the absolute tem-
perature.
Some additional arbitrary parameters included in the model are necessary
for numerical computations and some are included to allow facilities for
obtaining absolute values of the microstructure elements. These are: X
= the number of sets of polymer chains in the critical region; H = the
time increment for the numerical integration; Xmin= the minimum length
chain set in the critical region distribution at time zero; W = the total
chains/ per square centimeter in the flaw region at time zero.
The value of X and used were 101 sets and 1 sec, respectively; however,
variations of these values were checked to be certain that.the computer
solutions were relatively insensitive to the values chosen. Theoretically a
Gaussian distribution has no cut-off point. However, 99.74% of the distri-
bution lies within f35 of the mean value. This f3 s value was used as the
Xminpoint; however, the Xminlength value chosen at this point was purely
arbitrary. The total number of chains/cm2 in the flaw region was also an
arbitrary choice. Even so, in order to obtain chain scission results that
could be compared with experimental observations, an “in-theballpark”
number was necessary. No absolute values of the theoretical parameter
are known, and from an objective analysis one senses that the values that
could be selected encompass a fairly broad spectrum. There is some, of
course, interaction between the various model parameters and where one
is incorrectly chosen, other parameters must also be chosen in error to help
compensate or obtain a reasonable fit. I n fact, it was found for any one
type of experimental loading, a broad range of parameters could be chosen
and adjusted such that a reasonable theoretical fit could beobtained. How-
ever, when these same parameters were used to predict other loadings, an
extremely poor or inadequate fit was obtained for the majority of parameter
choices. From these observations it was concluded that the only valid
parameters would be those that successfully predicted the experimental
results under several types of loadings. As a result of this observation the
basic approach for determination of the parameters was as follows. Those
1424 LLOYD, DEVRIES, AND WILLIAMS

parameters that were best known, either from experiment or theory, would
be selected as base values. This included the distribution standard devia-
tion value (S) obtained from EPR data, the modulus of elasticity (Eb) for
a single nylon 6 chain from Treloarlmand the rate-theory collision param-
eter (wo),from Zhurkov and Toma~hevskii.~~ The other four parameters,
Re, RL, Uo, and y, were then selected by comparison of the theoretical
solution by using various combinations of the parameters with experi-
mentally observed data from constant strain rate and constant stress
(creep) tests of nylon 6 #1 at room temperature. Four average experi-
mental curves (two strain rate and two creep) were selected as the compari-
son standards. With these guidelines, parameter values at room tempera-
tures were then selected and the theoretical results plotted and compared
with the experimental standards. To obtain the “best fit” of the experi-
mental standards over 150 different combinations of the parameters, RL,
Re, Uo,and y, were computed and plotted as well as some variations of the
parameters, S, Eb, and WO. After considerable analysis a group of param-
eters was selected as the ‘(bestfit” with the knowledge that moderate varia-
tions would not substantially degrade and possibly would improve the
comparison with the experimental standards of nylon 6 #1 at room tempera-
ture. Once these room temperature parameters were determined for nylon
6 #l,they were used to predict behavior at other strain rates, creep, con-
stant stress rate, stress relaxation, and several frequencies of cyclic stress
fatigue. A very satisfactory correlation between experimental results and
theoretical predictions from the model was found and.have been presented.22
Table I summarizes the “best fit” parameters obtained at room tempera-
ture for nylon 6 #1 fibers.
The previous paperz2 discusses in some detail the parameters used in
this model. A pleasing aspect of the model was how nearly the parameters
which could be successfully used to predict behavior compared with ac-
cepted or theoretical values from other related studies. The parameters
used in the theoretical calculations could be varied moderately without
substantially degrading the correlation with some of the experimental
results. However, the range of variability was substantially reduced if

TABLE I
Model parameter Nylon 6 #1 “best fit” value
S, % strains 1.25
Rc 30
RL 5
Eb, lb/in.’ 3 x 107
WO, see’’ 10’J
UO,kcal/mole 67.5
Y %a 5
W , chains/cmab 1 x 101s
* Obtained from EPR data and varies with temperature and material.
b Arbitrary in theoretical model.
FRACTURE BEHAVIOR IN NYLON 6 1425

one was to correlate the model and experimental results for a variety of
different loadings, i.e., constant load rate, constant strain rate, creep, cyclic
loading, stress relaxation, etc. The creep and stress relaxation curves
proved to be most sensitive to variations in these parameters. The authors
were rather surprised as to the amount of stress relaxation that must be
attributed to bond rupture at high strain levels. The results of these stress
relaxation tests, extension of the model to other nylon fibers, and limits on
the model parameters will be presented in a future paper.
TEMPERATURE EFFECTS
Distribution
As pointed out above the key to understanding many of the experi-
mentally-observed properties of these highly-oriented fibers seems to be a
knowledge of the distribution of stresses in the polymer microstructure.
Theoretical calculations of the strength of a single polymer chain result in
tensile strength values of approximately 10,OOO ksi. However, tensile
strengths of most of the highly oriented polymer fibers rarely exceeds 150
ksi. This large difference has been explained as a consequence of the in-
homogeneity of the micromorphology in the fiber structure. Chain ends,
dislocations, disorder, disorientation, chain folds, and number and length of
tie molecules have all been suggested as elements that prevent a uniform
stress distribution in the fiber and result in regions of high stress concentra-
tion which in turn lower the maximum attainable tensile strength.2,11,17'32'33
Characterizing this distribution from EPR step strain tests and
relating it to the tie chains in the critical flaw regions turned out to be the
key result in the success of this study.22 Further step-strain tests at differ-
ent temperatures and on different nylon fiber samples were undertaken to
observe the influence of the effective distribution on the ultimate strength
of the fibers tested. These tests include: two sets of nylon 6 samples
(nylon 6 #1 and nylon 6 #2) at temperatures of -25"C, room temperature
(+22"C), +50"C, and -25°C; room temperature, +50"C, +75"C, and
+ lOO"C, respectively. A room-temperature test of a modified slack an-
nealed sample* of nylon 6 #2 and a nylon 66 sample at room temperature
was also made. Figure 3 shows typical results of these tests plotted as
histograms for different temperatures for sample material #2.
The effect of the width of the distribution on the maximum stress in these
polymer fibers can be observed in Figure 4. The interesting result from
this plot is that a direct correlation appears to exist between the width of the
distribution and the ultimate strength independent of the type of nylon
fiber, i.e., as the width of the distribution increases, the maximum strength
of the fiber decreases.
One can gain some insight into this effect by considering the extremes.
If the chains were all equally stressed, correspondingto a standard deviation
*Prepared by Dr. Joon Bu Park. Sample annealed with no restricting tension
at 2OO0Cfor 2 min and slowly cooled to room temperature before the stepstrain test.
1426 LLOYD, DEVRIES, AND WILLIAMS

120 120

u)

f40
a
2
0
0 4 8 I2 16
STRAIN (96)
a. b.

-: 80
m
3
\
u)
z
p 40
v)
Q

0
0 4 8 I2 16 0 4 8 I2 16
S T R A I N (ssl STRAIN 0
C. d.
Figs. 3. Histograms of samples of nylon 6, #2,from EPR stepstrain data at various tem-
+
peratures: ( a ) -25°C;( b ) room temperature (+22”C);(c) +50”C;( d ) 100OC.

of zero, the strength of the fiber would be the “strength” of each chain
multiplied by the number of chains across the critical cross-section of the
sample. Alternately, one might envision such a wide distribution in chains
that one chain is stressed and ruptured at a time. I n this case the strength
of the fiber reduces to the “strength” of the individual chains although the
energy dissipated during fracture and the strain at fracture are both greatly
increased. Real fibers behave more nearly like the former than the latter
of these extremes but it is rather amazing the very large differences that
small changes in the standard deviation of chain length can have on fiber
FRACTURE BEHAVIOR IN NYLON 6 1427

=cn
““1
180-

-
Y

In
ffl
W
[r
I-
cn
x 140-
Q
z

.8 1.0 1.2 1.4 1.6 I .8 2.0 2.2


I STO.DEV. ( x STRAIN)
Fig. 4. Effect of “apparent or effective” distribution on ultimate stength of nylon
fibers: ( I ) nylon 6, #1, -25°C; ( 8 ) nylon 6, #l room temperature; (5)nylon 6, #1,
+50”C; ( 4 ) nylon 6,# 2, -25°C; (6) nylon 6, #2, room temperature; (6) nylon 6, #2,
50°C; (7) nylon 6, #2, +75OC; (8) nylon 6, #2, 100°C; (9) nylon 6, #2 annealed at 200°C
and tested at room temperature;(10)nylon 66, room temperature.

strength. Although the correlation between strength and distribution is


evident from Figure 4, other factors such as the number of chains play an
important role in this relationship and are discussed later in this paper.
It should be noted that the term “effective or apparent distribution” has
been used for describing the experimental observations. For instance,
note the apparent narrowing of the distribution below room temperature
and the broadening above room temperature in Figures 3 and 4. This
change in the apparent or effective distribution with temperature gives im-
portant insight into the properties of the microstructure and merits a more
detailed discussion. Polymer chains in the flaw region are in varying states
of configuration and movement depending on several factors such as inter-
molecular forces, free volume, steric hinderances, and thermal environment.
The effect of temperature changes on the local surroundings can be thought
of as effectively tightening or loosening the structure. Tie chains are un-
doubtedly not rigidly embedded in the “crystalline block” material and as
a result, depending on the temperature or loading rate, some may pull out
(or partially pull out) rather than rupture. Alternately or in conjunction
with this effect, tie chains in kinked, coiled, folded, or other nonextended
configurations will most likely exhibit some relaxation, flow, unfolding, or
unkinking if aided by an increase in temperature. Likewise, with decreas-
ing temperature chains will tend to break prior to annihilation of kinks,
1428 LLOYD, DEVRIES, AND WILLIAMS

a. b.

C. d.
Fig. 5. Schematic representation of possible changes in effective tie chain distribution
with temperature: ( a ) unstressed control sample; ( b ) stressed sample at low tempera-
ture; ( c ) stressed sample at room temperature;( d ) stressed sample at high temperature.

coils, or folds and then the effective length of the chain may be much differ-
ent than the extended length. Therefore, a decrease temperature or ti high
loading rate should shift part of the chain length distribution to a smaller
effectivelength and a narrower distribution. The opposite effect would be
expected for higher temperatures. This effect is verified by the experi-
mental data in Figures 3 and 4. Figure 5 is a rough representation of
one of these concepts. The original unstressed tie chains are shown in
Figure 5a. At low temperature some of the chains are ‘(frozen” and have
smaller and more uniform effective length, resulting in a narrow distribution
(Fig. 5b). As the temperature increases, more unfolding or pulling out
takes place, resulting in a broader effective distribution as shown in Figures
5c and 5d.
These essentially reversible effects are not to be confused with irreversible
annealing effects which take place rather rapidly above 100°C. An ex-
haustive study has just been completed by Park under the direction of Stat-
ton on the effects of “stretch” and “slack” annealing on nylon 6 fibers.”
The slack annealed sample shown as point 9 in Figure 4 exhibited some
important differences to those pointed out above. Besides exhibiting a
.broader distribution and lower strength at room temperature than the con-
trol sample (point 5 in Figure 4), the distribution was shifted along the
strain axis to larger strain values. Results from Parkll indicate that during
slack annealing the polymer chains tend to become more disoriented and
chain folding occurs which may tend to pull some tie chains into the ordered
regions, decreasing the total number and relaxing the stresses in the tie-
chain regions. Figure 6 is a rough schematic representation of a proposed
FRACTURE BEHAVIOR IN NYLON 6 1429

a.

chain end

b.

C.
Fig. 6. Schematic representation of irreversible type changes created in the tie-chain
distribution under slack and tension annealing: (a) unstressed control; ( b ) unstressed
slack-annealedsample; (c) unstressedtension-annealedsample.

type of irreversible effects that “slack” annealing and “stretch” annealing


may produce in these regions relative to the control sample. From Figure
6b one can visualize a shift along the strain axis due to the overall effective
length increase in the tie molecules and a broader distribution. The lower
strength may be due to two effects: fewer tie molecules and less uniform
effective chain lengths. I n general, one would not expect a direct correla-
tion between the “effective distribution” and the maximum strength,
independent of the nylon type or treatment unless the total number of tie
chains was approximately the same. For example, if sample 1 has three
tie chains of equal length and sample 2 has six tie chains of equal length,
1430 LLOYD, DEVRIES, AND WILLIAMS

then both might have the same distribution, but sample 2 would be twice
as strong as sample 1. From the above reasoning the results of Figure 4
need to be analyzed in terms of the theory to determine the contribution of
each of the above effects to the observed experimental results. This will be
discussed later. Although “stretch” annealed samples were not tested and
characterized by these step strain tests, Figure 6c indicates the effect of this
process on the tie molecules as explained by Park. l1 Essentially some chains
may be fractured in the “stretch” annealing process but in general the
chains become better oriented thus narrowing the distribution and increas-
ing the strength. Further work characterizing “stretch” and “slack”

-I I

-8 1.0 1.2 I .4 1.6 I .8 2.0 2.2 2.4


I STD. DEV. (% STRAIN)

Fig. 7. Relationship between temperature and effective distributionfor various sample


materials: ( 0 ) nylon 6, #l; ( E l ) nylon 6, # l ; (A) nylon 6, #2 annealed (0)nylon 66.

annealed distribution properties will be published in the future. Figure 7


shows a plot of distribution width versus temperature for the nylon samples
tested. These sets of curves may be thought of as depicting the degree of
uniform chain orientation in the critical flaw regions relative to the type of
nylon. For instance, at room temperature nylon 6 #1 exhibits a narrower
distribution (more chains of the same effective length) than does nylon #2
at the same temperature. Likewise nylon 6 #2 at -25°C would be ex-
pected to exhibit similar properties to nylon 6 #2 at room temperature as
these effective distributions are similar. I n general, nylon 6 #1 would be
expected to be more brittle and stronger than nylon 6 #2 at the same tem-
peratures. Similar shaped curves would be expected for the slack-annealed
nylon 6 and for the nylon 66 samples but shifted as shown by the room
temperature values.
FRACTURE BEHAVIOR IN NYLON 6 1431

On analyzing the above results, the importance of the tie chain distribu-
tion in the critical flaw regions is readily evident, and any model which is
proposed to explain the failure mechanism in highly-oriented fibers should
include this distribution as a parameter.
Distribution and Temperature
As previously indicated, the authors have proposed a reaction-rate model
incorporating such a tie-chain distribution and successfully correlated room-
temperature tests for nylon 6 fibers. With the success of the criteria in ex-
plaining behavior at room temperature, it was thought wise to extend the
investigation to other temperatures. While the study has not been in great
depth, it did assess trends and provide insight into limitations and capabili-
ties of the proposed fracture criterion.
I n essence, kinetic rate theory states that the rate at which fracture of
bonds occur depends on three main factors: the activation barrier UO,the
thermal environment T , and the stress on the polymer chain, ui, i.e.
CB,= [WOexp - ((UO- rud/kT} ]Cut (7)
where C B i is the rate of chain scission, Curis the number of unbroken chains
in the ith family, wo,Uo,and y are kinetic parameters called the collision
parameter the activation energy, and the activation volume, respectively.
The stress on the bonds in the ith family is ut, T is the absolute tempera-
ture, and k is Boltzmann’s constant.
An increase in temperature aids the process in overcomingthe barrier and
increasing the stress effectively lowers the barrier. I n this light one might
say that the “strength” of a chemical bond at low temperature is greater
than that at high temperature.* From the observations reported here,
temperature has another effect on macroscopic strength. Increased tem-
peratures tend to loosen the structure and in this way broaden the effective
distribution. As a result the disparity in stresses on the polymeric chains
is increased and the resulting macroscopic strength is further decreased.
Figure 4 shows the relationship between the width of distribution and the
ultimate stress for several types of nylon at various temperatures. It
should be noted that experimentallytwo temperature effects were observed.
(1) At sufficiently high temperatures irreversible structural changes occur.
It has been observed that these could result in either broadening or narrow-
ing of the distribution depending on the presences of mechanical stresses
during the thermal treatment”. (2) A reversible loosening or tightening
of the structure effectively broadens or narrows the distribution, depending
on the sign of the temperature change.
It appears that both the decrease in bond strength due to thermal activa-
tion aiding bond scission and the broadening of the distribution due to
thermally induced loosening of the structure have a significant effect on
* It should be emphasized that “strength” is a relative term, since time plays such an
important role in the temperature-fracture behavior of polymers. Strength here im-
plies stress for a given lifetime.
1432 LLOYD, DEVRIES, AND WILLIAMS

macroscopic strength. In nylon 6 these two effectsare roughly comparable


as indicated in Figure 8. The solid lines in plots a and b of this figure show
the relationship between the experimentally observed maximum macro-
scopic stress and the experimentally determined effective distribution at
the indicated temperatures. The dashed line in Figure 8a shows the pre-
dicted effect of the distribution broadening alone. This curve was calcu-
lated by assuming room temperature strength of the chemical bonds, i.e.,

I60 a-25 ' C


-
v,
- 140-
Y

v)
v)
W
120'-
UJ

X
: 100-

80 I
1.0 1.2 14 1.6 1.8 2.0 2.2
a.

160-
-z
2 140-
u)
v)
W

:1 2 0 -
v)

X
a
E 100-

80 I
1.0 1.2 1.4 1.6 1.8 2.0 2.2
I S T D DEV ( X S T R A I N )
b.
Fig.8. Effect of apparent distribution on maximum strength for nylon 6, #2, at several
temperatures; (-) experiment; (a)dashed line, theory with distribution compensation
but without temperature compensation; ( b ) dashed line, theory with both distribution
and temptkature compensation.

that the rate of bond rupture at a given stress was what it would be at
room temperature. The dashed line in Figure 8b, on the other hand, tic-
counts for the increased rate of bond rupture as themal activation aids in
the bond degradation as well as the structural loosening effect described
above.
FRACTURE BEHAVIOR IN NYLON 6 1433

c-
-
ln
Y

v) 140-
ln
W
a
!-
v) +\

X
a
E

loo I 1 -100 -50 0 50 100 150


a
I8O 1

loo I
A
-100 -50 0 50
TEMPERATURE ("C)
100 150

b
Fig. 9. Effect of temperature on maximum strength for nylon 6, #2: - experiment;
(a)dashed line, theory with temperature compensation but without distribution com-
pensation; ( b ) dashed line, theory with both temperature and distribution compensation.

Note that at the higher temperatures in Figure 8b the theoretical strength


exceeds the experimentally observed strength. Although the difference is
not large (less than lo%), it is significant. Previous workn has indicated
that at temperatures approaching or in excess of the glass-transition tem-
perature, the mode of fracture changes and the tendency for the structure to
unfold or chains to pull out increases at a faster rate than the tendency for
backbone scission. In destroying the structure there is, therefore, a com-
petition between chain pull out or unfolding and bond rupture. The au-
thors suspect that for the nylon 6 fibers tested, as the temperature increases,
some of the tie chains will reach a cross-over point where pull out or unfold-
ing will occur under less stress than chain scission. Among other factors
this point varies, of course, with the length of chain embedded and, there-
fore, will vary from chain to chain. Hou.ever, at the point where a given
FRACTURE BEHAVIOR IN NYLON 6 1435

chain begins to pull out rather than rupture, the contribution of that chain
to the bulk strength of the fiber would be only slightly less than that re-
quired for chain scission. The observed strength is, therefore, only slightly
lower than that predicted at the highest temperature in Figure 8.
The curves in Figure 9 are analogous to those of Figure 8 just discussed.
Figure 9a compares the experimental results (solid line) with the predicted
effect solely due to the temperature changes (dashed line) assuming there is
no thermal effects on chain stress distribution. Once again, good agree-
ment is not possible without account for both thermal effects. By varying
the activation energy, better agreement than that of Figure 9a can be ob-
tained. However, the prediction would still be represented by a straight
line and in no way would be as good as that shown in Figure 9b. Similar
results were obtained on both types of nylon 6 (nylon 6 #1 and nylon 6 #2).29
Each material, however, exhibited a different distribution and, consequently,
a different strength. The thermal effects discussed here were identical in
concept for both and were equally good with regard to agreement between
experiment and predictions using the same kinetic parameters.
These pleasing correlations indicate the importance of a failure theory
which accounts for two major effects on an atomic scale to predict macro-
scopic behavior. Both effects are the result of the thermal environment
but one might be considered as an intramolecular effect while the other is
an intermolecular effect.
Two factors that are important in some polymers have not been included
in the analysis. (1) The model assumes that the polymer chains carry no
load until they become taut; it is known, however, that sizable "entropic
forces" can be transmitted by chains which are not taut; it has, in fact, been
demonstrated by P e ~ h h o l dthat
~ ~ the early part of the stress-strain curve
can be reliably reproduced by careful analysis of such forces. It will be
noted that for all the comparisons between the prediction of the model and
experiment presented in this and the previous paper, the correlation is less
good at low stress levels; undoubtedly, inclusion of entropic forces would
improve the agreement in this region. (2) Viscoelasticeffects resulting in a
timevarying stress on the polymeric chains have also not been included
per se. Such effects are undoubtedly present, but the good agreement ob-
tained while neglecting them suggests that they cannot be extremely sig-
nificant, at least in highly oriented nylon. To some extent viscoelastic ef-
fects are included as an ad hoc artifact in the kinetic parameters and the
distribution trends described above.
Figure 10 is a comparison between model predictions and experiment for
constant strain rate tests. Two different strain rates at each of three
temperatures (- 25"C, room temperature, +50°C) were theoretically pre-
dicted and experimentally observed. Seveml very important observations
can be made from the curves in this figure. Consider again the rate equa-
tion that was selected as the criterion for predicting chain scission, i.e.,
C m = [woexp { - (u,- y a J / k ~ 1) Cut (7)
1436 LLOYD, DEVRIES, AND WILLIAMS

If only changes in temperature (T)are considered, eq. (7) indicates that at


low temperatures the rate of chain scission should be less than a t high tem-
peratures. Therefore, if the effective distribution does not change with
temperature, the EPR curve for -25°C should shift to later times on the
time axis relative to the room temperature curve and the +50"C EPR curve
should shift to earlier times. Figure 10a shows this relationship as pre-
dicted theoretically by the model for -25"C, room temperature, and
+50"C, where the effective distribution was assumed to remain unchanged
from the experimentally observed room temperature value. Note also in
Figure 10a that the stress peaks occur in magnitude and order as +50°C,
room temperature, and -25"C, where that at +50"C occurs first in time
and is lowest in maximum stress. Comparing the predictions of Figure
10a with the experimental observations shown in Figure lOc, we note the
order of the predicted and experimental curves are reversed with respect to
each other in both the stress-peak sequence and the EPR curve sequence.
We conclude, therefore, that even reaction rate theory which incorporates a
distribution of stresses fails to predict these experimental temperature test
results if no compensation for a changing distribution with temperature is
included in the model. As the effective distribution changes with tempera-
ture, it could shift in several ways, two of which would be about its mean
value or about one end. Figure 11 shows the effect that these two types of
distribution shifts would have on the relative position of the EPR chain
scission curves for three temperatures. I n Figure l l a the relative position
of the curves is shown, where the effective distribution is assumed not to
change with temperature. Figure l l b shows the relative positions where
the effective distribution is changed about the mean value and Figure l l c
shows the relative values where the effective distribution change occurs
about the minimum length end. A comparison of these results with the
experimental observations in Figure 1Oc shows that the effective distribu-
tion changes with temperature by shifting about the minimum (or most
taut) chains length in the critical flaw region. Figure 10b indicates the
results of the theoretical predictions when the effective distribution is as-
sumed to shift about the minimum length as the temperature is changed.
This might be visualized by noting that in the critical regions, chains have
an original structure which undoubtedly includes kinks, coils, folds, or
other nonuniform length configurations. The shortest chains in this group
are most likely extended or taut between the two adjacent "cry-blk" re-
gions. As the sample is loaded and the region becomes extended, each
chain assumes some effective end-to-end length according to the effects of
temperature and/or strain rate on the deformation of the microstructure.
Therefore, at low temperatures the chains shift to shorter effective lengths,
resulting in a shift of the distribution mean toward the minimum chain
length side of the distribution. Therefore, the changing distribution and
changing temperature have opposite effects on the position of the chain
scission curves relative to the time axis (a two-fold mechanism), and the
macroscopically observed experimental EPR chain scission results are a
FRACTURE BEHAVIOR IN NYLON 6 1437

TIME

a.

A R
-25°C

21 ///
I
50°C R T -25'C

TI ME

b.

TIME

C.
Fig. 11. Schematic representation of relative EPR results with respect to distribution
changes with temperature: ( a ) no distribution change with temperature; (b) distribu-
tion shifts about its mean value as temperature changes; (e) distribution shifts about its
minimum length end as temperature changes.

compromise of the relative effects of both processes. Another observation


from the data shown in Figure 10 is the relationship (or lack of relationship)
between the total number of tie chains in the critical region and the number
of tie chains observed at the catastrophic fracture point. The fracture
point in each case was determined from the experimental results. The
point occurred at a specific time increment beyond the peak or maximum
stress of each test. The number of spins* at the fracture point and at the
maximum or peak stress point for nylon 6 #1 are plotted as a function of
temperature in Figures 12a and 12b, respectively. The theoretical values
in Figure 12 were plotted from the corresponding predictions in Figure lob.
Several observations can be made from the comparisons. The EPR values
recorded at fracture are dependent on how far beyond the maximum stress
point the test proceeds before catastrophic failure occurs. As the EPR
curve a t this point appears almost exponential for constant strain rate tests
* The authors prefer to defer until further studies and a latter paper a discussion of
the exact relationship between the number of spins (detectable free radicals) and the
number of ruptured bonds. For our present purposes it is important only that the num-
ber of spins be proportional to the number of bonds ruptured. The absolute number is
of secondary importance. There are, in fact, some aspects of fracture that would be
easier to explain if this constant of proportionality were rather high.
1438 LLOYD, DEVRIES, AND WILLIAMS

) 0
-100 -50 0 50 100 150
TEMPERATURE PC)
a.

lo t

-I00 -50 0 50 100 150


TEMPERATURE PC)

0.
Fig. 12. Effect of temperature o n total chain scission for nylon 6, #1: ( a ) at the
fracture point and ( b ) at the maximum stress point; (--)theory spins/cm2;(-) ex-
periment, spins/cma.

(Fig. lo), small time differences in the catastrophic fracture point can make
large differencesin the total number of chain scissions observed by the EPR
spectrometer. I n fact, the more narrow the distribution, the sooner cata-
strophic failure will generally take place, and therefore EPR results based
on the fracture point may appear significantly smaller. For instance, a
comparison of the differences between the experimental spin values at
-25°C and room temperature for the fracture point values of Figure 12a
and the maximum stress values of Figure 12b indicates a 30% increase in
relative spin values at the fracture point over the values a t the maximum
stress point. This increase can be attributed to differences in the time in-
crement to fracture after the maximum stress point was reached. Another
observation can be made from Figure 12b, i.e., the experimental number of
chain scissions increases with increasing temperature. This increase is also
noted in Figure 12a and at first glance might be attributed to small differ-
FRACTURE BEHAVIOR IN NYLON 6 1439

01 I , I I , I t

0 0.4 0.0 12 I -6 2.0 2.4


I STD.. DEV. 19)
Fig. 13. Theoretical maximum stress versus distribution width (per cent strain) curve
at room temperature for one specific set of model input parameters having an Rc value
of 30: (--) theory; ( I ) , (8), (S), ( 4 ) room temperature experiment values for nylon 6
#1,nylon 6 #2, nylon 6 #2 annealed, and nylon 66, respectively.

ences in the fracture point which could result in the large spin value varia-
tions. Apparently, however, only a portion of this increase can be attrib-
uted to such a process, as the results in Figure 12b also show this increasing
trend. Although the reason for this trend is not obvious to the authors,
it is important to note that the theoretical model also predicts an increase
in chain scissions at the maximum stress point as the temperature increases.
This theoretical relationship is obtained even though the total number of tie
molecules in the theoretical critical region remained constant. The addi-
tional differences in the experimental and theoretical values are most likely
related to the lack of including some viscoelastic effects as previously dis-
cussed as well as the experimental uncertainties involved in reading values
from the rapidly increasing portions of the curves. It is, therefore, con-
cluded from these observations that the total number of tie molecules in
this particular nylon 6 fiber have not sustantially changed during tests in
this temperature range. At higher temperatures some changes in the num-
ber of tie chain scissions might be expected as pull-out and unfolding occur
as discussed earlier. Thus, even though the total number of tie molecules
may remain constant, the number observed by EPR a t fracture or at max-
imum stress may vary depending on the distribution, temperature, strain
rate, and other factors.
Additional strain rate tests for nylon 6 #2 were experimentally and theo-
retically compared at three temperatures, -25"C, room temperature, and
+50"C. The results from these results were similar in every way to those
for nylon 6 #1 and once again substantiated the necessity for both bond
strength and distribution compensation at different temperatures.
1440 LLOYD, DEVRIES, AND WILLIAMS

Another interesting facet of the current study was the theoretical maxi-
mum strength that might be obtained from a nylon 6 fiber characterized at
room temperature by a specific set of parameters if only the width of the
effective distribution was varied. Such a theoretical curve is plotted in
Figure 13 using the “best fit” parameters for nylon 6 #1 and varying only
the effective distribution width. This particular curve is for a “crpblk”
to flaw region chain number ratio (Rc) of thirty. The maximum stress
values for a particular distribution width can be observed from this curve.
Note that nylon 6 #1 and nylon 6 #2 fibers both fall on this curve a t room
temperature but at different distribution widths. This reinforces previous
observations of the basic similarities between the two nylon fibers with the
exception of the “as received” distribution uniformity. Also, note that
the nylon 6 #2 annealed fiber and the nylon 66 fiber fall near but not on the
(‘best fit” parameter room temperature curve, indicating some slight basic
differences. The theoretical maximum strength value for a distribution
of zero width, i.e., all tie molecules of equal length, is predicted to be ap-
proximately 340 ksi. Noting that the assumed number of tie molecules is
thirty times less than the number that might possibly occur in a perfect
structure of infinite chain length, we obtain a maximum possible stress
value of 10,200 ksi for a perfect structure. This value is, of course, very
‘closeto the theoretical value of the strength of a single polymer chain.
From Figure 13 it is noted that at room temperature the annealed nylon
6 #2 fiber and the nylon 66 fiber values do not fall on the “best fit” curve as
determined from the nylon 6 #1 parameters. As the kinetic rate constants
(i.e., activation energy, activation volume, and collision parameter) would
not be expected to change for these materials, this observation indicates
basic change in the microstructure, i.e., the number, length, and distribution
of the tie molecules in the critical flaw regions. A discussion of the nylon 66
material will be deferred to a subsequent paper; however, a few comments
concerning the annealed nylon 6 #2 material will be made here. Testing
of the slack-annealed fibers by the authors was limited to room temperature
step-strain tests; however, in a related study additional testing was con-
ducted by Park.”
Combining experimental observations of Park” obtained from x-ray data
and noting the position of this sample in Figure 13, several observations can
be made. During slack annealing there is an increase in chain folding as
observed from x-ray data.” This would result in a slight decrease of tie
molecules in the flaw regions and a possibly larger increase in the chain ratio
parameter Rc, i.e., one tie chain could fold into several effective “cry-blk”
chains. Such an effect would not only account for the experimentally ob-
served lower than “best fit” theoretical values, observed in Figure 13, but
also would satisfy the total observed chain scission data for the annealed
sample, i.e., the total experimentally observed chain scission data for the
annealed sample was slightly less than that observed for the control sample.
Another effect of the slack-annealing process is that it apparently redis-
tributes the tie molecules more nonuniformly and perhaps relaxes to replace
FRACTURE BEHAVIOR IN NYLON 6 1441

the shortest chains in the region by a slightly longer set. These observa-
tions result from the apparent broadening of the experimentally observed
distribution and the shift of this distribution to higher strain values. Fur-
ther studies of both slack-annealed and tension-annealed samples by use of
the proposed theory would be most enlightening.

DISCUSSION AND CONCLUSIONS

It is interesting to speculate on the nature of the fracture mechanism.


Fracture is a complex phenomena and it is doubtful that EPR observes the
complete process. Macroscopic failure in these fibers might logically be
divided into three stages of development: (a) fracture initiation, (b) slow
crack growth of a large number of niicrocracks; (c) rapid crack propagation
of a few of the microcracks. The stages are characterized by different
processes. Fracture initiation describes the period from the beginning
of the mechanical excitation of the sample to the formation of a crack which
is large enough to influence its own growth rate. In these high-strength
fibers, apparently certain potential flaw sites are created during manufac-
turing, drawing, special treatments, etc. which can act as fracture-initiation
sites. These sites are distributed throughout the fiber and are composed,
in part, of tie molecules linking the “crystalline-like” regions together.
The relative number and arrangement of these tie chains determines which
sites might be considered “critical” sites. I n the first two stages of fracture
initiation many sites must be involved to explain the observed number and
distribution of radicals along the sample length. One would expect that
eventually one (or a few) of these critical sites will reach an unstable (Grif-
fith) size, and catastrophic failure of the fiber will rapidly ensue. This
latter stage will occur so rapidly and be so localized as not to be observable
by EPR. EPR then allows us to observe the development of microcracks
a t the critical sites during the first two stages. It is in the average of these
critical sites that chain scission is predicted by the theoretical model. The
model can apparently tell us little of the final stage. It was a rather pleasant
surprise how well an understanding of these two stages correlated with final
ultimate failure of the sample. This would seem to imply that in these
fibers at least these first two processes are the critical processes that
govern failure.
I n summary, the goal of this study was to establish a microscopic failure
criterion for highly ordered semicrystalline polymers in fiber form which
would relate microscopic occurrences to macroscopic behavior. The study
was limited to a narrow and yet very important class of materials, the
highly oriented semicrystalline polymer fibers, specifically nylon 6, where
failure is suspected to be dominated by polymer chain scission. The
mathematical model is an idealized model of the polymer fiber micro-
structure, representing mainly how the structure acts during the failure
process and undoubtedly neglects many morphological intricacies.
1442 LLOYD, DEVRIES, AND WILLIAMS

Uniaxial tensile loadings were the only modes used in testing the fibers.
Loading rates were relatively high and time-to-failure short in order to
minimize viscoelastic and free-radical decay effects.
Viscoelastic time effects and entropic forces were not included in the
mathematical model; however, portions of the curves attributed to this
latter effect22have been treated with some success by another
Several major observations on the basic model are in order. (1) Re-
action-rate theory with a stress-aided activation energy is a successful
failure criterion for predicting covalent bond rupture under mechanically
induced stress. The key to the success of the theory lies in the proper
interpretation and application in the stress u. The stress u is the stress
on an individual bond and must be properly related to the macroscopic
applied loadings by the micromacro structure relationships. (2) The
mathematical model provides direct verification of the need for a distribu-
tion in length of tie molecules to explain experimentally observed strength
and other fracture-related properties of these nylon fibers. EPR provided
a method of experimentally characterizing the “apparent or effective”
distribution due to temperature, annealing treatments, or manufacturing
processes.
A puzzling observation in the earlier EPR studies here and elsewhere has
been the variation in number of free radicals produced at fracture at dif-
ferent temperatures and strain rates. The combined effects of thermal
activation, the distribution in stress among the chains and thermal loosen-
ing of structure quantitatively explain these effects. It is in fact observed
that the overall effects of varying temperature on the mechanical behavior
of fibers requires an understanding of variance in chain-load distribution.
Results at different temperatures further substantiate the importance
of the distribution in stress among the polymer chains in describing me-
chanical behavior using reaction rate theory. Increasing the temperature
has two related but separate effects. First, thermal activation aids back-
bone scission, therefore, effectively decreasing the covalent “bond strength”
at higher temperatures. This increases the rate of bond rupture at a given
bond stress (force). Second, temperature serves to loosen the polymer
structure by altering the mechanical forces required for chain unfolding,
unkinking, untangling, and breaking of van der Waal bonds for slip, etc.
This serves to alter the stress distribution which once again effects the
fiber “strength .”
To account for the effect of stress on the macroscopic strength of the
fibers, one must account for both these factors. I n this study if either factor
n-as neglected in the analysis neither qualitative nor quantitative agree-
ment with experiment was obtained.
It is possible to explain the difference in strength between different
samples of nylon 6 by the differences in their microstructure. The most
important factors apparently being how well the chains are aligned, the
number of tie chains, and how equally the stress is distributed among the
tic chains in the critical regions. (A factor deserving further study, in-
FRACTURE BEHAVIOR IN NYLON 6 1443

cidentally, is the size and distribution of these critical regions and means
by which they might be controlled and/or modified.)
The key to improving the properties of polymeric fibers, therefore, lies
in modifying the alignment and effective length and/or number of tie
chains. Park, Statton, and D e V r i e ~have~ ~ been rather successful in de-
veloping stretch-annealing treatments for use in bringing about such modi-
fications. The techniques and models outlined herein provide a systematic
means of analyzing the effect of such treatments.
I n closing, a few comments on macroscopic nonuniformity of the sample
material is perhaps in order. As previously discussed,22there is, of course,
a possibility that the “apparent” or effective distribution could be due in
part to macroscopic differences in fiber lengths during the manufacturing
of the filament windings or during sample preparation. There are, how-
ever, at least two factors that support the microscopic explanation. (1)
A one standard deviation of 10% of the average value would mean a total
apparent distribution (including 99% of all fibers) of about 30a/c. The
samples had an initial length of 2s cm and, hence, to account for the dis-
tribution from purely macroscopic effects overall differences of +S cm in
length would be required. Considerable care was taken in sample prepara-
tion to ensure equal length of the yarns in a sample bundle and while there
was undoubtedly differences in the fiber lengths in the samples, these would
not be expected to exceed a few tenths of a centimeter. (2) Further evi-
dence of a microscopic distribution is obtained from tests by Verma and
Peterlinlg on oriented nylon films and a limited number of studies in this
laboratory on creep of highly-drawn monofilament nylon 66 rods. Qualita-
tively, at least, these results are similar to those for the studies reported
herein. As noted previ~usly,’~ theories not incorporating a stress distribu-
tion due to the material microstructure have been very unsuccessful in
predicting bond rupture kinetics for various loadings.
Besides differences in length of the fibers within a sample bundle, there
could also be other physical differences between the fibers in a yarn.
P r e ~ o r s e khas
~ ~ recently noted some quasi-macroscopic differences in
fibers within a yarn that closely relates to this discussion. He noted an
apparent difference in the extent of drawing within a yarn although each of
the fibers should have nominally received the same treatment. Dr.
Prevorsek has reported the birefringent data shown in Table I1 on a yarn

TABLE I1
An Example of Macroscopic Nonuniformity:
Data of Birefringence Distribution of a Nylon 6 Yarn*
Sample size 92
Average birefringence 0.0545
Standard deviation 0.0016
Lowest value 0.0508
Highest value 0.0597
8 Data of Prevorsek.“
1444 LLOYD, DEVRIES, AND WILLIAMS

of nylon 6. The extent to which this variation would effect the individual
stress-strain bond-rupture response of the fibers in the yarn is not pres-
ently known. However, one might expect the difference in orientation for
example as manifest by these birefringent differences to macroscopically
distribute the load among the different fibers. I n the present model there
is no way to distinguish this from the microscopic effect.
A supporting argument that a significant amount of the effective dis-
tribution is due to microscopic changes in the structure is the fact that this
effective distribution changes with temperature. It is difficult to see how
such changes with temperature can be attributed entirely to macroscopic or
quasi-microscopic distribution effects which may occur during the fiber
manufacturing process or during sample bundle preparation. These
latter effects are fixed once the process is complete and would not change
with temperature unless the microstructure changes.
It is hoped that a study now in the planning stages in connection with
Dr. Prevorsek will help determine what portion the quasi-macroscopie
effects play in the observed effective distribution.
Grateful acknowledgment is given to the National Science Foundation for their por-
tion of the financial support of this work. Use WBS made of facilities purchased under a
National Aeronautics and Space Administration grant. Thanks are expressed to Dr.
Joon Bu Park, Dr. William 0. Statton, Dr. A. Peterlin, and Dr. H. H. Kausch for their
interest and suggestions. Gratitude is also expressed to Mr. Stephan Nichols for help
in accumulating and reducing the data.

References
1. J. W. S. Hearle and R. H. Peters, Fiber Structure, The Textile Institute, Manches-
ter, 1963.
2. A. Peterlin, J. Polym. Sci. A-2, 7, 1151 (1969).
3. R. Bonart and R. Hosemann, 2.Elektrochem., 64,314 (1960).
4. J. W. S. Hearle, in Supramolecular Stmcture in Fibers ( J . Polym. Sci. C, 20),
P. H. Lindenmeyer, Ed., Interscience, New York, 1967, p. 215.
5. D. E. Bosely, in Supramolecular Structure in Fibers ( J . Polym. Sci. C, 20), P. H.
Lindenmeyer, Ed., Interscience, New York, 1967, p. 77.
6. A. Keller, Phil. Mag., 2 , 1171 (1957).
7. E. W. Fischer, H. Goddar, and G. F. Schmidt, Makromol. Chem., 117,170 (1968).
8. P. F. Dismore and W. 0. Statton, J. Polym. Sci. B, 2 , 1113 (1964); Small Angle
Smttering From Fibrous and Partially Ordered Systems ( J . Polym. Sci. C, 13), R. H.
Marchessault, Ed., Interscience, New York, 1966, p. 133.
9. A. I. Gubanov and A. D. Chevychelov, Soviet Phys. Solid State, 4 , 4 (1962).
10. M. Takayanagi, K. Imada, and T. Kajiyama, in U.S.-Japan Seminar in Polymer
Physics ( J . Polym. Sci. C, 15), R. S. Skin and S. Onogi, Ed., Interscience, New York,
1966, p. 263.
11. J. B. Park, Ph.D. Dissertation, Division of Materials Science and Engineering,
Department of Mechanical Engineering, University of Utah, Salt Lake City, Utah,
1971.
12. A. Tobolsky and H. Eyring, J . Chem. Phys., 11,125 (1943).
13. F. Bueche, Physical Properties of Polymers, Wiley, New York, 1961.
14. A. I. Gubanov and A. D. Chevychelov, Soviet Phys. Solid State, 5,62 (1963).
15. B. Rosen, Fracture Processes in Polymer Solids, Interscience, New York, 1964.
FRACTURE BEHAVIOR IN NYLON 6 1445

16. S. N. Zhurkov, A. Y. Savostin, and E. E. Tomashevskii, Soviet Phys. Dokl., 9,


968 (1964).
17. D. K. Roylance, K. L. DeVries, and M. L. Williams, Fracture, P. Pratt, Ed.,
Chapman and Hall, London, 1969.
18. H. H. Kausch and J. Becht, Rhwl. Acta, 9,137 (1970).
19. G. S. P. Verma and A. Peterlin, Polym. Preprints. 10, No. 2, 1051 (1969).
20. H. H. Kausch, Znt. J. Fracture Mech., 6,301, (1970).
21. H. H. Kausch, J. Macromol. Sci. Rev. C4,243 (1970).
22. K. L. DeVries, B. A. Lloyd, and M. L. Williams, J. Appl. Phys., 42,4644 (1971).
23. J. Becht and H. Fischer, Kolloid-Z., 229,167 (1969).
24. D. K. Roylance, Ph.D. dissertation, Department of Mechanical Engineering,
University of Utah, Salt Lake City, Utah, 1968.
25. J. Becht, K. L. DeVries, and H. H. Kausch, Europ. Polym. J.,7,105 (1971).
26. D. K. Backman, Ph.D. dissertation, Department of Mechanical Engineering,
University of Utah, 1969.
27. D. K. Backman and K. L. DeVries, J. Polym. Sci. A-1,7,2125 (1969).
28. A. D. Chevychelov, Polym. Mech., 3 , 5 (1970).
29. B. A. Lloyd, Ph.D. dissertation, Department of Mechanical Engineering, Uni-
versity of Utah, Salt Lake City, Utah, 1972.
30. L. R. G. Treloar, Polymer, 1 , 9 5 (1960).
31. S. N. Zhurkov and E. E. Tomashevskii, Proceedings of the Conferenceon the Physi-
cal Basis of Yield and Fracture, Oxford University Press, Oxford, 1966.
32. P. Predecki and W. 0.Statton, J. Appl. Phys., 37,4053 (1966).
33. K. E. Perepelkin, Polym. Mech., 2,536 (1969).
34. W. Pechhold, in Molecular Order-Molecular Motion ( J . Polym. Sci. C, 32).
H. H. Kausch, Ed., Interscience, New York, 1971, p. 123.
35. J. B. Park, W. 0. Statton, and K. L. DeVries, submitted to Journal of Polymer
Science.
36. D. C. Prevorsek, personal communication.
Received March 17, 1972

Vous aimerez peut-être aussi