Introduction

Porosity is an intrinsic property of the polyacrylonitrile (PAN) fibers that is due to their microstructures and exists as voids (pores) between the lamellae, crystals and other elements of the fibers’ texture [1, 2]. Their mechanical properties are largely limited by the presence of the pore defects where the stress accumulates, accelerating the formation and the development of cracks, which eventually results in catastrophic break [3, 4].

At present, there are many spinning techniques to make PAN precursor fibers, including wet spinning, dry spinning, and melt and electrospinning. However, to date, the dry spinning and melt spinning have yet not been reached the carbon fiber qualities by wet spinning, since they bring many defects into the fibers [5]. Meanwhile, in electrospinning, the fibers are short and are collected in the form of fiber mats rather than a continuous fiber filament [5, 6]. Therefore, the traditional wet spinning is the preferred technique for manufacturing carbon fiber precursors. Furthermore, to reduce the processing cost, the dry-jet wet spinning with high speed is developed to prepare high-quality precursor fibers [7]. During fiber formation and conversion process, the morphologies and size distribution of the pores in fibers were adjusted through controlling the spinning conditions [8]. Therefore, to decrease fiber porosity and increase fiber mechanical properties, it is important to study the evolution of pore structures and the relation between pores and mechanical properties during the dry-jet wet spinning process.

The pore structures in PAN fibers are originally formed in the coagulation bath during the spinning process [9]. The spinning dope solutions including dimethyl sulphoxide (DMSO) solvent and PAN were extruded into DMSO/H2O coagulation bath and then became unstable thermodynamically due to counter diffusion phenomenon. Furthermore, the phase separation occurred and a polymer-rich phase formed the fiber skeletons, while a solvent-rich phase formed pores [10]. Therefore, the counter diffusion process was affected by variation in effective parameters of spinning dope and coagulation bath conditions, which had a significant influence on pore formation and growth [10, 11]. During post-spinning process, the DMSO solvent was gradually removed, resulting in shrinking of fiber skeletons and the partial collapsing of the pores [12]. Therefore, the pore development was controlled by drawing conditions.

In conclusion, there are two effective methods to improve the size and number of the voids and reduce the overall porosity of PAN fibers during the spinning process, namely preventing voids growth in the coagulation bath and healing the previously formed voids by hot drawing the nascent fibers [11, 13, 14]. Unfortunately, the pores are hard to eliminate completely. Therefore, the pore defects may still retain their geometrical identity through the manufacturing process and would generate into the final carbon fibers, which are thought to be responsible for the decrease in carbon fiber tensile strength [2, 15]. Accordingly, it is necessary to acquire more information on emerging pores during manufacturing process, thereby minimizing the porosity of the PAN precursor fibers through control of the spinning conditions.

The pore structures of the PAN fibers have been studied by several methods to estimate their parameters. The most common indirect methods including the classical absorption/desorption [16, 17], mercury porosimetry [1, 18], and thermoporometry [9, 12, 19] are frequently used, with their underlying principle including the effect of capillarity and fluid flow in the porous material or on the filtration effect [20], which only provide the dimension information. However, the most direct method of the electron microscopy makes it possible to reveal the shape and morphology of pores through the evaluations of the microscopic pictures of the fiber with image analysis.

Atomic force microscopy (AFM) can only focus on nanopores of the fiber surface [2], and the pores with relatively larger diameters distributed in the fiber surface and cross section are easily observed by scanning electron microscopy (SEM) [9, 11, 21, 22]. However, it is hard to obtain the morphology and information on the location of the pores in fiber’s interior directly. In addition, the method of preparing the objects also restricts exploring the pores. To determine the fine structures, high-resolution transmission electron microscopy (HRTEM) is an indispensable method [23, 24] and the ultrathin section is also a sufficiently well technique for interior fine structure observation [23, 25].

According to the International Union of Pure and Applied Chemistry (IUPAC) recommendations, there are three kinds of pores dimensionally: micropore (D < 2 nm), mesopore (2 nm < D < 50 nm) and macropore (D > 50 nm). The micropores in PAN fibers are considered as sufficiently compact, since the distance between the elements is very small [26]. However, mesopores that distinctly perturb the compactness and homogeneity of the PAN and carbon fibers have a negative effect on their mechanical properties and are difficult to observe by direct electron microscopy. Herein, longitudinal ultrathin sections of PAN fibers were prepared and the mesopore structures were characterized by HRTEM. In addition, the frequency histograms for mesopore sizes were also presented and analyzed. The effect of the coagulation conditions and hot-drawing conditions on the mesopore variation is discussed. A possible correlation between the mesopore structures of PAN fibers and their mechanical properties is proposed.

Experimental

Preparation of PAN fiber

The PAN copolymer was obtained through an aqueous polymerization in our laboratory. The acrylonitirle (AN, AR) and itaconic acid (IA, AR) in water were initiated by ammonium persulfate (APS, AR) to prepare the PAN copolymer (AN:IA = 98.5:1.5) with a viscosity molecular weight of approximately 1.5 × 105 g/mol (measured by the viscometry, as shown in Fig. S1, in Supplementary Materials), under the condition of nitrogen protection. Then, PAN powders were deposited on the bottom of the flask being the only component insoluble in water. Finally, the PAN powders and DMSO solvent with the mass ratio of 18.5:81.5 were mixed by the machine of a double-screw extruder, yielding the spinning solution dope.

The dry-jet wet spinning process was conducted on a pilot plant at Shandong University, China, as shown in Fig. 1. The spinning dope was extruded under the pressure through the spinneret (3000 orifices, D = 0.12 mm diameter) and passed through the coagulation bath (30 wt% DMSO solvent and 70 wt% water, 5 °C) to obtain the PAN nascent fibers designated as: N1, N2 and N3 samples. The parameters of these nascent fibers are listed in Table 1. The air gap distance for the nascent fibers N1, N2 and N3 were 3 mm, 8 mm and 3 mm, respectively, while the drawing ratio (defined as the ratio of the first roller take-up velocity V1 to the dope extrusion velocity V0) was set as 2.5, 2.5 and 3.8, respectively (keeping V1 constant and different drawing ratios could be attained by changing V0). Therefore, the time of the counter diffusion was the same in all nascent fibers. Then, the nascent fiber N3 was stretched in multi-steps in warm water with degrees of temperature (ranging from 45 to 80 °C) with a total drawing ratio of 2.9 to obtain PAN fiber P1. Moreover, the PAN fibers were immersed in oil for isolation and protection, and successively densified at 140 °C with a drawing ratio of 1 to obtain PAN fiber P2. Finally, the PAN fibers were stretched in water vapor with a drawing ratio of 3.5 to obtain PAN fiber P3 (precursor fiber). In order to assess the mesopore structures during the spinning process, some PAN fiber samples (nascent fibers N1–N3 and PAN fibers P1–P3) were collected (as shown in Fig. 1; Table 1) and treated further.

Fig. 1
figure 1

Dry-jet wet spinning process and the PAN fiber samples collected point

Full size image
Table 1 Parameters of the nascent fiber and sampling point of the PAN fiber
Full size table

Preparation of the ultrathin section

The fiber object for HRTEM observation was prepared by an ultrathin sectioning technology. Firstly, some fibers were embedded and fixed by epoxy resin solution to prevent fiber sliding over cutting process. Following solidification, longitudinal ultrathin sections were cut through a diamond knife by an ultratome (LKB-2088, BROMMA Ltd., Sweden). The sections were controlled within 50 nm in thickness. Finally, the ultrathin sections were easy to float on the surface of purified water in the sink. To better support samples, copper grids with supported carbon film were used to collect ultrathin sections.

Characterization

The longitudinal ultrathin sections of PAN fiber samples were examined by HRTEM (JEM-2100, JEOL Ltd., Japan) at an acceleration voltage of 200 kV. The image processing technique was carried out to measure the pore size of the mesopore structures in fiber samples using the ImageJ software. According to the results given in Figs. S2 and S3 (ref to Supplementary Materials), only the thickness contrast made a contribution to the electronic imaging during characterizing the PAN fiber longitudinal ultrathin section by HRTEM. Therefore, the white region in HRTEM image was considered as the space of the mesopore structures in the fiber. The width (perpendicular to the fiber axis) and length (parallel to the fiber axis) of the mesopores were proposed in this paper to characterize the fiber size property. Furthermore, the pore size distribution was represented by a histogram describing the frequency of pore size:

$${\text{Frequency }}\left( \% \right)\,=\,{\text{number}}\,{\text{of}}\,{\text{pores}}\,{\text{with}}\,{\text{pore}}\,{\text{size}}/{\text{total}}\,{\text{number}}\,{\text{of}}\,{\text{pores}}.$$

The frequency histogram pictures for width and length of the mesopores are also presented as analyzed by the Origin software.

The tensile strengths of fiber samples were determined by a monofilament tensile testing machine (XQ-1, Donghua University, China) at room temperature. During the measurement, the gauge length and stretching speed were chosen as 20 mm and 10 mm/min, respectively. A series of 10 single-fiber samples were carried out for the tensile test measurement.

Results and discussion

Mesopore in the PAN nascent fiber

Figure 2 shows the HRTEM morphology of the PAN nascent fiber longitudinal ultrathin section. The morphologies of surface region and core region of PAN fibers are presented in Fig. 2a–f, respectively, taken from a single fiber. All the HRTEM images show a lamella-like structure perpendicular to the fiber axis, formed by folded molecular chains during the coagulation process [27,28,29]. In addition, the mesopore structures are distributed between the lamellae.

Fig. 2
figure 2

HRTEM morphology of the PAN nascent fiber longitudinal ultrathin section a, bN1; c, dN2; and e, fN3

Full size image

Compared to wet spinning where fiber filaments were immersed directly in a coagulation bath followed by their instant coagulation, in dry-jet wet spinning process, a thin solidified skin (as shown in Fig. 2) was built to avoid sudden mass transfer by diffusion [30]. Meanwhile, the molecular orientation in nascent fiber induced by shear stress within the spinneret could be frozen into the wet-spun fiber but relax in a small air gap region for the dry-jet wet-spinning fiber [31]. During the coagulation process, a phase-separation occurred due to solvent/non-solvent exchange, which was restricted by a thin solidified skin. Meanwhile, the molecular chains were folded into the lamellae [29] and a poor-polymer phase was formed into the mesopores distributed between the lamellae. The mesopores might be considered as the path where DMSO solvent is diffused out along the fiber transverse direction. Consequently, the counter diffusion was promoted by not only the concentration gradient but also by the compression field in the transverse direction.

Figure 2a, b presents the lamellae in the nascent fiber N1. There are many mesopores between the lamellae and they have the relatively larger number and length in the core region than those in the surface region. On increasing the air gap, the more lamellae structures and oriented structures in the surface region were frozen due to significant solidification by cooling or gelation by moisture evaporation on fiber surface, which increased the thickness of the skin. Meanwhile, relaxing time also increases, which induces more oriented chains in the core region to relax.

Therefore, the lamellae in the surface region of the nascent fiber N2 (Fig. 2c) are packed more closely than that of N1, while the lamellae are arranged loosely and a large number of mesopores are distributed unevenly in the core region (Fig. 2d). Furthermore, compared to the nascent fiber N1, the morphological difference between the surface region and the core region became more obvious for the N2. The orientation and regularity of the molecular chains in N3 are improved by increasing the drawing ratio. As shown in Fig. 2e, f, the thickness of the lamellae in N3 is smaller than that in N1. Furthermore, the morphological difference between the surface region and core region of N3 is narrowed, and the number and size of the mesopores are decreased.

The dimensions of the mesopores perpendicular (width) and parallel (length) to the fiber axis were determined by measuring the linear distance of the white region in the respective direction. The histograms in Fig. 3 show the frequency distributions of width and length of the mesopores in the nascent fibers. The distributions of the mesopore length in the surface region are unimodal, but other distributions of the mesopores are multimodal. Compared with the nascent fiber N1, the length and width distribution are relatively broad in N2. Contrarily, the length and width distribution become narrow in N3 by increasing the drawing ratio.

Fig. 3
figure 3

Histograms of the size distribution of mesopores in nascent fiber

Full size image

The numerical values of mesopore width and length of the nascent fiber are given in Table 2 according to Figs. 2 and 3. In all nascent fibers, the mesopore length values in the core region are larger than those in the surface region, but the widths are complex. A significant difference in the shape of mesopores has appeared in the nascent fibers, in particular that of the aspect ratio (α = length/width). The aspect ratios of the mesopores of nascent fiber N1 are significantly small with α = 0.66 and α = 0.84, indicating that these mesopores belong to the linear channel pores and their major axes are perpendicular to the fiber axis. On increasing the air gap, the aspect ratios of the mesopores in the surface region and core region of the nascent fiber N2 are α = 0.78 and α = 1.19, respectively. Consequently, the relaxation of the oriented molecular chains in the long air gap may bring a huge effect on the shape of the mesopores. For the nascent fiber N3, the mesopores have an almost round shape with aspect ratios close to 1 and the mesopore sizes decrease apparently.

Table 2 Mesopores size of nascent fibers
Full size table

A measure of the mesopore area is the product of width and length, which is defined as effective mesopore area Aeff [2], and the effective mesopore area of fiber longitudinal ultrathin section can be used to represent fiber volume. The effective mesopore areas in the surface and core region of the nascent fibers N1, N2 and N3 amounted to Aeff = 368.15/357.00, Aeff = 488.89/382.95, and Aeff = 264.04/266.59, respectively. Compared to the nascent fiber N1, the mesopore volume of N2 is increased and the mesopore volume of N3 is decreased.

During coagulation process, the formation and growth of the mesopores were influenced by the variation in drawing ratio and air gap distance. By increasing the air gap, the diffusion of solvent and non-solvent was limited by increasing the skin thickness, which would bring unwanted influence on the mesopore structures, and the mesopore sizes and volume increased in N2. As for N3, a high drawing ratio helped alignment of the molecular chains and tended to push the solvent out and prevent the entry of the non-solvent into the core during the coagulation process [32]. Therefore, the diffusion of the solvent/non-solvent is decreased by increasing the drawing ratio, which could be attributed to the orientation of the voids along the fiber axis, the increase in dense ligament and the loss of connectivity between the voids (blocking transport channels of solvent and non-solvent) [10, 33, 34]. Thus, the sizes and volumes of the mesopores are decreased in nascent fiber N3.

Mesopore structures/mechanical properties relation in PAN nascent fibers

The mechanical properties of the prepared PAN fibers are listed in Table 3. Due to relatively large titers of nascent fibers, nascent fibers N1, N2 and N3 have smaller tensile strength values [7, 35, 36]. There are no obvious grooves and cracks on the surface of nascent fiber, as shown in Fig. S4 (in Supplementary Materials). Thus, according to the Griffith’s theory [37], the fiber tensile strength is mainly limited by the mesopore defects content, where the stress is accumulated and the crack is formed easily under load, resulting in fiber’s catastrophic failure. Furthermore, the comparison of the mesopore dimension with mechanical properties of different nascent fiber samples reveals a clear correlation, as demonstrated in the correlation plot in Fig. 4. It shows that the tensile strength of the nascent fiber decreases with the increase in mesopore width and length. This lowering trend in tensile strength with increasing effective mesopore area may be caused by higher number of large defects and the breaking original fibers.

Table 3 Mechanical properties of PAN nascent fibers
Full size table
Fig. 4
figure 4

Correlation between tensile strength and mesopore dimensions of the nascent fibers

Full size image

Mesopore in the PAN fiber during spinning process

A large number of pore defects were existed in the nascent fibers, which were the main cause of the lowered tensile strength. Therefore, there was a chance to reduce the porosity of the PAN fibers through hot drawing at various stages of the post-spinning process, such as drawing in hot water washing bath, in hot steam chambers or on the hot rollers [5, 13, 22]. However, drawability of the PAN fibers during the post-spinning process was determined by the size distribution and number of the mesopores and interior microstructure morphologies of the nascent fibers. Consequently, the nascent fiber N3 was chosen to prepare the PAN precursor fibers, experienced the process of washing, densifying and steam stretching.

Figure 5 shows HRTEM morphology of the PAN fiber longitudinal ultrathin section. Compared to the nascent fibers, the lamellae are packed more closely in the surface region but there is still a great deal of mesopores in fiber P1 (Fig. 5a, b), resulting in DMSO removed and relaxing the lamellae ordered structures in the washing bath. The mesopores with oval shapes are oriented along the fiber axis under the stretching field. After densification treatment, the mesopores have been collapsed and merged. Consequently, the number and size of the mesopores have decreased in fiber P2 (Fig. 5c, d), but there is still a part of interior water failed to eliminate due to the dense skin, and some mesopores being merged to form a few macropores in the core region. After steam stretching, the lamellae are packed more closely in the fiber P3 (Fig. 5e, f) and the number and size of the mesopores are decreased.

Fig. 5
figure 5

HRTEM morphology of the PAN fiber longitudinal ultrathin section a, bP1; c, dP2; and e, fP3

Full size image

The histograms in Fig. 6 show the frequency distributions of width and length of the mesopores in the PAN fibers. The center of the length and width distribution is moved into smaller size during the spinning process. The histograms of the length and width of the mesopores show a symmetric frequency distribution, which may be caused by a small number of the mesopores with large sizes. Their length and width in the core region are larger than those in the surface region. In addition, there are some macropores in the core region of PAN fibers P1 and P2. Consequently, the size distribution of the mesopores shifts toward smaller values through closely arranged lamellae during hot-drawing process.

Fig. 6
figure 6

Histograms of the size distribution of the mesopores in PAN fiber

Full size image

The numerical values of the mesopore width and length of the PAN fiber are given in Table 4 according to Figs. 5 and 6. Apparently, the lengths of the mesopores in the PAN fiber are larger than their widths, and the mesopores have the oval shape with an aspect ratio > 1, oriented along the fiber axis. These results are attributed to the stretching field during post-spinning processing. The effective mesopore areas of the PAN fibers P1, P2 and P3 have amounted to Aeff = 193.42/ 298.87, Aeff = 133.09/262.22, and Aeff = 40.85/181.16, respectively. Compared with the nascent fiber N3, the mesopore volume of PAN fibers is decreased successively during the post-spinning process.

Table 4 Mesopores size of PAN fiber
Full size table

After coagulation process, the nascent fibers have been washed to remove residual solvent using warm water and led to high diffusion of water into the fiber, which could cause large voids. To prevent the formation and development of a large number of the mesopores in the washing bath, the several washing baths in different gradient temperatures were used to avoid rapid influx of water [5]. Furthermore, the use of drawing enhanced the axial orientation of the molecular chains and fiber linear density to decrease the mesopore volume. The mesopores in the washing steps were stretched along the fiber axis. Consequently, compared with the nascent fiber N3, Aeff in the surface region was decreased and Aeff in the core region increased instead in the PAN fiber P1, but obviously their aspect ratio has been increased. During the densification process, wet fibers were treated on the hot rollers with roller temperature being greater than the glass transition temperature of PAN fibers. Therefore, the water in pores was desorbed from the wet fibers, which led to the collapse of the mesopores. Consequently, the mesopore sizes decreased in the PAN fiber P2 and the Aff in the surface region and core region decreased by 31.2% and 12.2%, respectively. During the saturated steam stretching process, the water steam serving as a plasticizer could reduce the Van der Waals force between the molecular chains at high pressure and temperature condition, which could improve the molecular orientation along the fiber axis and decreased the size and volume of the mesopores [38]. Therefore, the mesopore sizes decreased obviously in PAN fiber P3 and Aff in the surface region and core region by 69.3% and 30.9%, respectively. Consequently, the steam stretching was more effective in reducing the size and volume of the mesopores in PAN fibers in comparison with their drawing in washing bath and densification process.

Correlation between mesopore structure and mechanical properties of PAN fiber

The mechanical properties of the PAN fibers are listed in Table 5. During the post-spinning process, the titers of the PAN fibers decrease (Titer: P1 > P2 > P3) and their tensile strengths increase (tensile strength: P1 < P2 < P3). Consequently, the hot drawing can efficiently improve the tensile strength of the PAN fibers. As shown in Fig. S5 (in Supplementary Materials), the surfaces are smooth and there are no obvious grooves and cracks for PAN fibers. So, decreasing the content and sizes of the mesopore defects is one of the most efficient ways to enhance the tensile strength of the PAN fibers from the point of the material defects [39]. Comparison of the mesopore dimension with mechanical properties of different PAN fibers reveals a clear correlation, as demonstrated in the correlation plot in Fig. 7. It shows that the tensile strength of the PAN fibers increases with decreasing mesopores width and length. Consequently, the content of the mesopores decreases by the hot drawing during the post-spinning process, which can improve the tensile strength of the PAN fibers.

Table 5 Mechanical properties of the PAN fibers
Full size table
Fig. 7
figure 7

Correlation between tensile strength and mesopore dimensions of the PAN fibers

Full size image

Conclusion

The mesopore variations in PAN fibers during the dry-jet spinning process were investigated. The mesopores were mainly distributed between the lamellae in the PAN fibers. The mesopore sizes and volume of the nascent fibers increased with increasing the air gap and decreased with increasing the drawing ratio. Their widths were larger than their lengths. Then, the size and volume of the mesopores decreased through hot drawing during the post-spinning process. Their lengths were larger than their widths, which indicated that the mesopores were oriented along the fiber axis under the stretching field. The mesopore sizes and their amount in the core region were larger than those in the surface region, indicating that the denser surfaced structures and loose core part were coexisted in PAN fiber, termed as “skin-core” structures. Due to the relatively large titer of the nascent fibers, they had smaller tensile strengths. The tensile strength of the PAN fiber samples increased during the post-spinning process. The mechanical properties of the PAN fibers were correlated with the dimension of the mesopores. The fiber tensile strength increased with decreasing the mesopore width and length for the nascent fiber and PAN fibers.