1. Introduction
High-performance oriented polymer materials are promising materials in various fields of application. One of the effective methods of their production is the drawing of isotropic low-strength polymers [
1]. The initial lamellar crystallites in isotropic polymers are transformed in the course of orientation into rigid oriented fibrils due to the unfolding of the polymer chain folds in the initial lamellae. Such a discrete “structural re-arrangement” leads to a significant (by an order of magnitude or more) increase in the material important mechanical characteristics —Young’s (or elastic) modulus (
E) up to 200 GPa and strength (
σ) up to 6 GPa [
1,
2,
3]. The advantage of such materials with respect to high-performance inorganic materials lies in their extremely high specific characteristics (related to the material density
ρ). For instance, for a high-strength steel characterizing with very high values of
σ = 1–2 GPa [
4], the
σ-to-
ρ ratio, at a steel density of
ρ ≈ 8 × 10
3 kg/m
3, is
σ/
ρ = 0.13–0.25 × 10
−3 GPa·m
3/kg while that for the ultra-oriented ultra-high-molecular-weight polyethylene (UHMWPE) gel fibers with
σ = 6 GPa [
2] and
ρ ≈ 1 × 10
3 kg/m
3 attains
σ/
ρ = 6 × 10
−3 GPa·m
3/kg, i.e., it is 25–50 times greater than that for steel.
However, one of the drawbacks of the drawing process is the material embrittlement resulting in a sharp increase in a scatter of the measured mechanical properties. This behavior is due to the formation and development of surface microcracks at the final stages of orientational drawing, the so-called kink bands, which are initiators of the most intense crack growth at the deformation stages preceding the sample failure [
1,
2]. For this reason, in order to correctly estimate the average values of
σ,
E, and strain at break (
εb),
σav,
Eav, and
εav, which are important characteristics playing a key role in choosing the most appropriate fields for material applications, the number of identical samples to be tested should be increased markedly, from the generally accepted about five [
5,
6,
7,
8] to some dozens, or even hundreds [
2,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22]. In this case, the statistical distributions of mechanical properties can also be analyzed, and additional information obtained, for instance, on their conformity to a certain theoretical model (e.g., Gaussian [
23], Weibull’s [
9,
10,
1112,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22]). It can be helpful for a better understanding of the deformation and fracture mechanisms of materials and for making a more proper recommendation about their practical use. Actually, if the distribution of a measured dataset can be represented with a bell curve that is characteristic of the Gaussian (or normal) distribution, it implies that the data scatter is caused by the sum of many independent and equally weighted factors [
23]. By contrast, if it obeys the Weibull’s statistics, it should have the shape of a linear plot in specific coordinates “lnln[1/(1 −
Pj)] − ln
σ”, where
Pj is the cumulative probability of failure, suggesting the dominant role of the surface cracks and their propagation across the sample cross-section in the sample fracture [
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12].
Earlier [
2,
3,
15], it has been shown for high-performance polymer materials that the more correct type of the statistical distributions, Weibull’s or Gaussian, of
σ,
E, and
εb, and the values of the statistical Weibull’s modulus (
m) characterizing the data scatter (an increase in
m means a decrease in the data scatter) were dependent both on the chain chemical structure and the sample type (single or multifilament fibers). However, statistical characterization of other mechanical properties, besides
σ,
E, and
εb, has not been investigated yet, though it is helpful for a better understanding of the deformation and fracture mechanisms. In this respect, it is interesting to perform a detailed analysis of the stress–strain curves of such materials in order to estimate other important characteristics [
24] in addition to those indicated above.
So, the goal of our study was to carry out a detailed analysis of the stress–strain behavior of a number of high-strength polymer materials in an attempt to estimate additional mechanical characteristics, besides widely used σ, εb, and E, and to investigate the conformity of their statistical distributions to the Gaussian and Weibull’s models.
For this purpose, UHMWPE, polyamide 6 (PA 6), and polypropylene (PP) were chosen. The choice of these polymers was motivated by different chain architectures and conformations resulting in their different abilities to strengthen. Each of these polymers has been investigated by using two different oriented sample types: single or multifilament fibers. It was motivated by different statistical natures of these samples: “statistical origin” of a multifilament sample (including some hundreds of individual fibers) and “non-statistical origin” of a single fiber. The stress–strain behavior of these high-strength oriented polymer materials will be analyzed in more detail than usual (i.e., analysis of σ, εb, and E only), and the statistical distributions of new mechanical characteristics estimated when analyzing the stress–strain curves will be compared with those of σ, εb, and E.
It should also be noted that, along with drawing, some enhancement of the mechanical properties of polymer-based materials can be obtained by introducing the filler into a polymer matrix (see, e.g., [
25,
26,
27]). However, despite the significant increase in the values of
σ and
E for such composite materials with respect to those of neat polymers, the former are still markedly lower (
σ = 0.7 GPa,
E = 17 GPa [
25]) as compared to those of highly oriented polymers, in particular, of gel-cast UHMWPE fibers (
σ = 6 GPa,
E = 200 GPa [
1,
2,
3]).
4. Results
Typical stress–strain curves for the high-performance single and multifilament fibers of PP, PA 6 and UHMWPE are shown in
Figure 1. One can see that the curves for PP and UHMWPE demonstrate a monotonic increase in
σ with
ε until the sample fracture while those for PA 6 have a steeper curve portion at
ε > 6% with respect to the initial curve portion at
ε < 1%, indicating that additional strengthening takes place in the course of sample deformation. This behavior suggests that the strengthening potential has not been exhausted in the course of the drawing process, and the mechanical properties can be enhanced markedly even after an additional drawing of the samples by 10% only.
Let us consider the stress–strain curve for the PA 6 multifilament for which this effect is more marked in more detail by taking the first derivative of the curve (see
Figure 2). In this case, one observes two tangent lines corresponding to the Young’s modulus (
E1) at small strains < 1% and an apparent viscoelastic modulus (
E2) at large strains (
ε > 10%). By extrapolating the second curve portion (
E2) to
σ = 0, the value of strain at break
εb for this
ε range (
εb-2) can be estimated. In this way, the second portion of the curve can be adjusted to a typical stress–strain behavior in the shape of a monotonic increase in
σ with
ε observed here for the PP and UHMWPE materials. By using this approach, one can also estimate, besides the ‘traditional’ mechanical properties such as
σ,
εb and
E1, the values of
E2 and
εb-2 and, thus, characterize the mechanical behavior of the PA 6 fibers more fully.
We first performed the statistical analysis of the characteristics which are relevant to all the six materials investigated, in particular,
σ; and then
E2 and
εb-2 found on the PA 6 stress–strain curves were considered. For this purpose, the
σ values for each of the materials were arranged in an ascending order (see
Figure 2a) and analyzed relative to their conformities to the Weibull’s and Gaussian statistical models. The results of this analysis are shown in
Figure 2b,
Figure 2c,
Figure 2d,
Figure 2e,
Figure 2f,
Figure 2g and
Figure 2h, respectively. In addition, the Weibull’s analysis results are collected in
Table 1. It is worth noting that the choice of the high-performance polymer materials under study allowed us to analyze their statistical mechanical behavior over a rather broad interval of strengths, from 0.2 to 6 GPa.
As follows from
Figure 2b and
Table 1, the constructed Weibull’s plots in the majority of cases (UHMWPE-m, PA 6-s, PA 6-m, PP-s, PP-m) represent single linear curves fitted with rather high reliability (root-mean-square deviation
R2 > 0.95). For the UHMWPE single fibers, the data can be correctly fitted with two linear curves. However, 80% of the data pointing at higher strengths also demonstrate a high linear fitting reliability (
R2 = 0.996). Besides, taking into consideration that the ratio of the parameter
σ0 estimated from the Weibull’s plots (having the physical meaning of
σav) to
σav,
σ0/
σav, is close to unity in all the cases considered, the results of the Weibull’s analysis performed seem to be correct.
As far as the Weibull’s modulus m is concerned, its value depends on: (i) the polymer’s chemical structure; (ii) the sample type; and (iii) the strength interval considered. In general, the m value for the multifilament is higher than the corresponding m value for the single fiber (carbon-chain UHMWPE and PP) or close to it (heterochain PA 6). For the ultra-high-strength UHMWPE single fibers, a markedly higher value of m = 74.0 was calculated for 20% of the samples with low strengths with respect to that of m = 7.7 calculated for 80% of the samples with high strengths. By contrast, a single linear curve with m = 9.9 is obtained when fitting the entire dataset of the UHMWPE multifilament. To put it differently, a higher m value for the UHMWPE multifilament (m = 9.9) as compared to that for the majority of the data for the UHMWPE single fiber (m = 7.7) indicates that the data scatter is narrower for the former. This observation suggests that the UHMWPE multifilament demonstrates a more reliable mechanical behavior as compared to that for the UHMWPE single fibers. Moreover, the observation of an excellent reproducibility (m = 74.0) of the low strength values (σ ≈ 4 GPa) for the latter ones indicates that the use of such materials is potentially more dangerous because there is an increased probability of fracture at σ < σav with respect to that for the UHMWPE multifilament. The fact that the m values for the UHMWPE and PP multifilament are higher with respect to the corresponding m values for the single fibers indicates that the role of critical cracks in the entire sample fracture is more important for the latter ones, which seems to be reasonable. Actually, the fracture of one filament should not necessarily result in the fracture of the entire multifilament sample including some tens or hundreds of single filaments.
The analysis of the strength data distributions on their conformity to the Gaussian model (see
Figure 2c–h) has revealed that, for five of the six materials investigated, except the UHMWPE single fibers (see
Figure 2g), the distribution histograms represent the bell-shaped curves suggesting that the measured data can be satisfactorily described in the framework of this model. It is also seen that, in general, (i) the fitting results are more correct for the three different multifilament samples than those for the corresponding single filaments, and (ii) for the materials with a lower strength.
When comparing the applicability of the Weibull’s and Gaussian models to the strength distributions of the high-performance polymer materials investigated, one can distinguish the statistical dualism of the strength distribution behavior for the materials with
σ < 1 GPa. To put it differently, the four datasets for the PA 6 and PP materials investigated have been found to obey both the Weibull’s and Gaussian statistics. The conformity to the Gaussian statistics may be explained, on one hand, by their (though rather limited) plasticity (
εb = 10–20%) preserving their brittle fracture. On the other hand, they do not undergo yielding (yield points are absent on the stress–strain curves—see
Figure 1a). So, they can also be treated as quasi-brittle materials for which the Weibull’s statistics were proposed initially. The UHMWPE materials investigated characterized with smaller values of
εb < 5% are referred to as quasi-brittle materials. So, it is not surprising that their strength distribution behaviors are in good accord with the Weibull’s model. The fact that the Gaussian model does not always work properly for these materials, in particular for the UHMWPE single fibers, indicates that the data
scatter is not caused by the sum of many independent and equally weighted factors [
23]. Rather, it is controlled by weakest links and a crucial role of surface cracks in the sample fracture, i.e., the key factors on which the Weibull’s statistics are based [
2,
21].
Let us turn to the statistical analysis of the stress–strain curves of the two PA 6 materials under study by comparing the distributions of
E1 (=
E) and
E2, and
εb and
εb-2. The results of this analysis are presented in
Figure 3 (
E1 and
E2 for single fibers),
Figure 4 (
E1 and
E2 for multifilament fibers),
Figure 5 (
εb and
εb-2 for single fibers), and
Figure 6 (
εb and
εb-2 for multifilament fibers), and are collected in
Table 2.
Let us apply the Weibull’s and Gaussian models to describe the statistical distributions of the values of
E2 and
εb-2 presented in
Figure 3a and
Figure 4a, and
Figure 5a and
Figure 6a, respectively, in the same way as above for
E1 and
εb. For this purpose, the values of
εb,
ε0,
E, and
E0 in Equations (7)–(10) should be replaced by the corresponding values of
εb-2,
ε2-0,
E2, and
E2-0. As a result, one obtains Equations (11)–(14) that will be used for the Weibull’s analysis of
E2 and
εb-2 below.
As follows from
Figure 3a, the values of
E1 and
E2 for the PA 6 single fibers are close, indicating that the effect of the deformation strengthening is not characteristic of this material. These data are represented in Weibull’s coordinates “lnln[1/(1 −
Pj)] − ln(
property)” in
Figure 3b. The Weibull’s modulus values calculated from the tangents of these plots
m(
E1) = 23.4 and
m(
E2) = 14.9 indicate that that the scatter data for
E2 are broader than that of
E1. The data analysis in the framework of the Gaussian model shows that the bell-shaped curves can be received by computer fitting of the histograms for both
E1 and
E2 (see
Figure 3c and
Figure 3d, respectively). However, the curve obtained for
E1 with a well-defined maximum is more symmetric while that for
E2 demonstrates a rather diffuse maximum shifted to higher
E2 values. Hence, the Gaussian model describes the distribution of
E1 more correctly as compared to that of
E2. By contrast, the Weibull’s model describes more correctly (
R2 = 0.936; for the
E1 distribution,
R2 = 0.907) the
E2 distribution. These two observations may be explained as follows. On one hand, the
E2 value is calculated for the strains > 0.5
εb. In this strain range, the onset of the local fracture processes (chain scission, interfibrillar slipping) is expected. Hence, this mechanical property ‘approaches’ the fracture properties such as
σ and
εb for which the Weibull’s model works properly [
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15]. On the other hand, the
E1 value characterizes the very beginning of the sample deformation. At this deformation stage, the contribution of the fracture processes such as covalent bond rupture and crack propagation is negligible. Therefore, the validity of the Weibull’s model in this case is less appropriate than the Gaussian one. Therefore, it may be concluded that the
E1 data scatter is controlled by many independent and equally weighted factors, as expected in the framework of the Gaussian model.
Let us consider variations of the same mechanical properties for the PA 6 multifilament fibers. For this material, the
E2 values are twice as high compared to the
E1 values (see
Figure 4a). This means that a significant improvement in the material rigidity takes place with an increase in
ε. The Weibull’s modulus values calculated from the plots of
Figure 4b are 17.2 and 49.6 for
E1 and
E2, respectively, i.e., the observed difference by a factor of three is rather large, indicating that the data scatter of
E2 is significantly smaller as compared to that of
E1. It may be indicative of the development of a more uniform structure in the course of deformation. As far as the Gaussian analysis is concerned, the histograms PDF(
E1) and PDF(
E2) presented in
Figure 4c and
Figure 4d, respectively, show that the
E1 distribution can be fitted satisfactorily with a bell curve. By contrast, two maxima with rather high PDF values of 40 and 30% are observed on the
E2 histograms, indicating that the
E2 distribution in the framework of the Gaussian mode is bimodal and, hence, it has a rather complicated character.
Let us analyze the distributions of
εb and
εb-2. The datasets
εb(
n) and
εb-2(
n) for the PA 6 single fibers presented in
Figure 5a are plotted in the Weibull’s coordinates in
Figure 5b. As one can see, the two linear fit curves are symbatic, and they are characterized with rather close
m values for
εb-2 and
εb (
m = 8.94 and 9.40, respectively). The analysis of these data using the Gaussian model (see
Figure 5c,d) shows that the results received are arbitrary. Therefore, the Weibull’s model seems to be more appropriate for describing correctly the statistical distributions of
εb-2 and
εb for the PA 6 single fibers in comparison with the Gaussian model. Taking into consideration that the Weibull’s model has been proposed for a fracture property (
σ), this behavior seems to be reasonable. For the PA 6 multifilament samples, one observes behavior similar to that for the PA 6 single fiber samples: symbatic curves
εb(
n) and
εb-2(
n) (see
Figure 6a) and close tangents of the linear fit curves lnln[1/(1 −
Pj)] = f(ln
εb) and lnln[1/(1 −
Pj)] = f(ln
εb-2):
m = 35.0 and 23.5, respectively.
It should be noted that the Weibull’s analysis carried out for the deformation characteristics considered is correct for the two sample types since it is characterized with R2 ≥ 0.98 and R2 ≥ 0.96 for the single and multifilament PA 6 fibers, respectively. However, in contrast to the Gaussian distributions of εb and εb-2 for the PA 6 single fibers, the fitting results for the PA 6 multifilament, in particular, the curve for εb-2, using the Gaussian model seem to be correct. The difference in the Gaussian distributions of εb and εb-2 revealed for the two PA 6 sample types may be explained by the statistical nature of one multifilament sample (including about 200 individual fibers) in comparison with one fiber in the single fiber sample. To put it differently, the test results for the multifilament sample are substantially more verified statistically which is important for obtaining a proper Gaussian distribution.
From the data collected in
Table 2, it follows that all the values of the ratio of the scale parameter to the average value of the corresponding mechanical property
p0/pav (
E2-0/E2-av,
ε2-0/ε2-av,
E1-0/Eav,
σ/
σav, and
ε0/εav) for the two PA 6 sample types investigated are close to unity (
p0/pav ≈ 1), indicating that the Weibull’s analysis carried out is correct for all five mechanical properties investigated. In order to estimate the effect of the deformation strengthening at the second portion of the stress–strain curve, let us compare the statistically correct values of
E1 and
E2, i.e., the values of
E1-0 and
E2-0, respectively, calculated from the corresponding Weibull’s plots. Since the ratios
E2-0/
E1-0 for the single and multifilament PA 6 fibers are 0.98 and 1.55, respectively, this means that the slope of the second portion (at
ε > 6%) of the stress–strain curve for the PA 6 multifilament is steeper than the initial portion (at
ε < 1%) while those for the single fibers are almost equal. This suggests a marked effectiveness of strengthening of the PA 6 multifilament. It should also be noted that the applicability of the Weibull’s model is more correct for a fracture property (
εb-2,
R2 > 0.96) with respect to a viscoelastic property (
E2,
R2 < 0.94). This behavior seems to be reasonable since this model has been proposed initially for a fracture property (
σ).
Finally, let us compare the
m values estimated for the two PA 6 sample types investigated (see
Table 3) as the
m ratios for the multifilament to those for the single fiber.
It is seen that these ratios are rather close for σ and, to a lesser extent, for E1 while they are markedly higher for the multifilament, by a factor of 3 to 4, for E2, εb, and εb-2. This means that, in general, the data scatter for the multifilament is substantially smaller than that for the single fiber. This behavior may be explained by an extremely large number of the fractured single fibers per one test for the multifilament sample including 200 single fibers with respect to the single filament sample (i.e., 200 against 1), making the results for the former more appropriate statistically which is reasonable.