1 Introduction

In liquid composite molding (LCM) processes, flow velocity is determined by the filling of micro and macro voids of the porous medium, which is commonly referred to as dual-scale flow [1,2,3,4]. A good balance between these mechanisms is key for void-free impregnation of textile reinforcements [1, 4,5,6,7,8] and consequently for a cost-effective process [9]. While the flow through macro voids is mainly influenced by the permeability of the textile, the flow through micro voids is dominated by capillary forces inside the fiber bundles. Although the influence of capillary flow is in most cases much smaller than that of the (injection-)pressure-driven, hydrodynamic flow, it should not be neglected for the filling of a dual-scale porous medium [4]. Micro voids inside the rovings caused by a rapid hydrodynamic flow are even more difficult to avoid than macro voids caused by a faster capillary flow [1].

Preforms for LCM can be made by textile processes, such as stitching, knitting, weaving, braiding, or sewing. Binder/tackifier materials are the chemical approach to produce preforms to provide stability, near-net-shape form, and handling abilities [10,11,12]. Often powdered, epoxy-based materials are used, which show thermoplastic behavior under certain processing conditions but can then be incorporated in the composite’s matrix after resin infusion or injection [10, 11].

Nevertheless, the binder particles can be considered as foreign material within the reinforcement structure and influence many aspects of the processing of the preforms. They have a significant impact on ‘springback’, respectively the compression and relaxation properties, of the stack [12,13,14,15,16]. Also, the final mechanical properties can be affected severely by the presence of a binder [17,18,19].

In case of LCM-manufacturing of a composite part, the influence on the resin flow needs to be investigated. Some studies have dealt with the influence of binder on the permeability in preforming processes such as dry fiber placement (DFP). In DFP a slight increase in permeability was found for a larger amount of binder [17]. Yet, particles blocking the flow channels in contrast to the possibility of enlarging (or even creating) flow channels and the reduction of hydrodynamic compaction during the infusion lead to ambiguous results [17]. In studies concerning preforms of weaves or non-crimp fabrics (NCF) also inconsistent results were reported [10,11,12, 14, 15, 18, 19]: The blocking or the creation of gaps due to binder particles as well as a varying melting behavior leads to either an increase or a decrease of permeability. Results vary especially for high activation temperatures. It can be stated that the molten binder with its high viscosity (compared to infusion resins) creates a film between the layers and is partially imbibed by the rovings due to capillary forces. Depending on the complex processing conditions and fabric structure, permeability is influenced differently.

Nevertheless, a key aspect here is the effect of the binder on the capillary tubes inside the rovings after its activation. Rohatgi and Lee [14] performed experiments with a centrifugal device to calculate the capillary pressure during drainage. They concluded that lower capillary pressure, with the binder imbibed by the roving at high activation temperature and pressure, could be a sign that flow into the tows was hindered by the binder.

This study aims to investigate whether the capillary imbibition of binder by the fabrics at different activation temperatures affects the capillary flow of the whole fabric stack. For this purpose, capillary rise experiments were performed. In these, capillary flow is often modeled in a first approach by the well-known Lucas-Washburn (LW) Eqs. [20, 21]:

$${z}^{2}\left(t\right)=\frac{\gamma \text{c}\text{o}\text{s}\left(\theta \right){r}_{c}}{2\eta }t$$
(1)

In this equation, the squared rise height depends on the surface energy \(\gamma\), the cosine of the contact angle \(\theta ,\) the capillary radius \({r}_{c}\), and the dynamic viscosity \(\eta\). All these parameters can potentially be influenced by additional material such as binder. The influence of the liquid and surface properties is neglected in these experiments due to the usage of the same test fluid and a similar test setup. The capillary radius (respectively the sum of the radii of all capillary tubes in a roving) is the parameter which is under investigation here.

2 Materials and experimental procedure

Four common carbon fiber textiles with epoxy-based binder applied were chosen for this study to investigate the capillary rise behavior after activation of the binder. Due to the availability and the favored use as part of (preformed) high-performance parts only carbon fiber fabrics were selected. Table 1 shows the used materials. While the two non-crimp fabrics (NCF) show a fine layer of binder coated on one side, the two woven materials are covered loosely with larger binder particles. Binder amounts vary between 10 and 20 g/m² with a large possible deviation according to the supplier’s datasheets. The chosen fabrics cover a large range of different textile structures, fiber areal weight (FAW), and binder appearances.

Table 1 Overview of the carbon fiber textiles used in the experiments
Full size table

The following experimental procedure is shown in Fig. 1. First, all specimens were cut on a Zünd G3 cutter (Zünd Systemtechnik AG, Switzerland) with the dimensions 60 × 65 mm. Due to material and machinery availability, two different approaches to activate the binder were investigated. Though they cannot be compared directly, they can be regarded as different methods to prepare a preform. Stacks of eight layers for the NCF +/- 45° fabric and six layers for the other materials were prepared directly after cutting. The NCF +/- 45° material was conditioned in an autoclave at the University of Nottingham with a pressure of 6 bar for one hour. Activation temperatures of 80, 100, 120, and 140 °C were used. The other materials were prepared in a vacuum bag on a heating plate, i.e. without the external pressure of the autoclave. Here, specimens with activation temperatures from 70 to 130 °C were prepared every 10 °C. After the heating plate reached the desired temperature, it was held for 30 min. For reference, stacks were kept in a vacuum bag at room temperature for the same duration.

Fig. 1
figure 1

Experimental procedure

Full size image

Capillary rise experiments have been performed on a test rig which has been presented in more detail in [22, 23]. In these experiments, the stacks are compressed between two glass blocks so that a planar flow will be established. A test fluid in a container on a vertical translation stage (in this study: n-decane) is then lifted until it comes into contact with the fibers. Capillary rise can then be observed by camera images of the flow front and synchronous recording of the weight of the imbibed fluid. A 3 mm cavity distance frame was used for the two materials with a lower FAW (Twill 2/2 and NCF 0°) and a 4 mm frame for the others. Table 2 shows the resulting fiber volume content (FVC).

Table 2 Calculated FVC of the stacks in the cavity
Full size table

Experiments were repeated 10 times for NCF +/- 45° and for the other materials five times. Temperature was recorded during the experiments to calculate the influence on the fluid parameters. It was at 22 °C +/- 2 °C for all experiments.

After the flow front evaluation, solely the capillary rise behavior of the NCF +/- 45° specimen showed a significant difference between binder at room temperature and at elevated temperatures. Therefore, the stack of this material was separated and the interlayer fracture surface was investigated with a digital microscope (VHX 7000 series, Keyence Corporation, Japan) at all the used temperatures. The fracture surfaces of the other fabrics were inspected after activation at 140 °C, ensuring full melting and distribution of the binder.

3 Results and Discussion

3.1 Capillary flow

In the capillary rise experiments, the flow front of the test fluid was evaluated via digital image processing. The empirical cumulative density function (eCDF) at 95% was used for an accurate estimation of the flow front. Figure 2 shows the results of these experiments. The depicted lines show the mean of all measurements per temperature enveloped by small dotted lines representing the standard deviation.

Specimens with binder activated in an autoclave (a) show different behavior compared to the specimens activated on a heating plate in a vacuum bag (b-d). Temperature levels of 80, 100, and 120 °C show a slower capillary rise behavior than the reference stacks which were kept at room temperature and have not been treated. A slightly faster rise than the other activated fabrics can be observed for the stacks prepared at 140 °C. It can be seen that binder activation decelerates the capillary flow. For the highest activation temperature of 140 °C, the flow velocity is faster again, but clearly slower than for the non-activated stacks. To investigate this behavior in more detail, digital microscopy images of the interlayer fracture surface were taken and will be discussed in the following subchapter.

The three materials activated on a heating plate (b-d) show inconsistent results in the capillary rise experiments, but some conclusions can be drawn nevertheless. In case of the weaves (b and d), all specimens are in a narrow range. They all are roughly within standard deviation. Inaccuracies originating in measurement and sample preparation have a larger influence than the binder material, which in these cases consists of large-sized particles. The unidirectional NCF (c) shows a difference between the specimens at room temperature and the activated ones. Activation of the binder accelerates capillary rise whereas the specific temperature is of less importance, similar to the weaves. With the molten binder creating a barrier between the layers the fluid is forced to stay in the capillary tubes. This blocking of peripheral flow accelerates the flow front at the beginning of the rise. With the progressed filling of the fabrics, this effect weakens at a certain height.

Fig. 2
figure 2

Overview of capillary rise experiments at room temperature: mean (thick lines) and standard deviation (dotted lines); Lucas Washburn (LW) fit added in a)

Full size image

A side observation of these experiments is that textiles with a higher fiber areal weight show a slower rise velocity but a more uniform behavior. A high fiber areal weight allows for fewer pores in the stack and additional compaction can even decrease the capillary radii and hence reduce flow velocity. The closure of large, irregularly distributed gaps however provides a more uniform flow.

It can be concluded that for preforming processes without additional external pressure, the influence of the activation temperature of the binder on capillary flow cannot be predicted reliably (much like the influence on the permeability). It appears that many opposing mechanisms act on capillary flow, although it is not clear which exerts the strongest influence during the measurement.

3.2 Digital Microscopy

To investigate the binder distribution of the NCF +/- 45° fabric at the various temperature levels digital microscopy was chosen. A magnification of 200x shows a good overview of a representative area of the textile’s surface. To investigate the structure of the binder between the layers the stacks were separated and the fracture surface was recorded.

Fig. 3
figure 3

Fracture surfaces of NCF +/- 45° with binder activated at various temperatures after separating the stack

Full size image

Figure 3 shows the binder on the fracture surface activated at four different temperature levels. According to the technical data sheet of the binder [24] the softening point is at 102 ± 5 °C. At 80 °C droplets of binder material are clearly visible. At 100 °C almost the whole surface is covered by the binder material. The morphology of the binder drops is still visible and the roving surface is partially recognizable. Temperature levels well above the indicated softening point lead to a complete wetting of the surface. The drop morphology becomes less visible with increasing temperature. Moreover, the structure of the fabrics below the binder surface becomes more apparent again. This can be explained by a very thin and uniform wetting by the binder material and/or imbibition of liquid binder into the fabric.

Additionally, imprints of the rovings from the adjacent layer can be observed on the fracture surface. While they are just some grooves on the resin drops at 80 °C, at 140 °C the whole area is covered with an imprint of the fibers. The differences in the images correspond well to the results of the capillary rise experiments (and with [18]).

Comparing the behavior of the NCF +/- 45° fabric to the others, they were also activated at 140 °C for 30 min in a vacuum bag. The left column in Fig. 4 shows the binder distribution on the surface of a single layer before activation. The middle column shows the surface of the layer that was exposed after removing the upper three layers of six. Magnifications of fiber imprints in the binder material of this layer can be seen in the right column.

Fig. 4
figure 4

Binder distribution of different fabrics on the surface before activation (left), on the fracture surface after activation at 140 °C and centered separation of the layers (middle) and magnification of representative areas with fiber imprints of the adjacent layer (right)

Full size image

NCF 0° shows a fine distribution of the binder particles which are relatively small in size. After activation, the presence of the binder is merely recognizable. A thin film is established, respectively the binder is absorbed into the rovings. Imprints of the neighboring layer can be seen predominantly by the sewing threads. They leave a uniform mark on the binder material. Binder particles on the satin and twill weave fabrics are larger and fewer in number. Activation leads to melting and local imbibition. The rovings are not fully covered with binder. Traces of the separation of the layers can be found for both materials close to binder accumulations where the fiber imprints can be spotted.

An explanation of the capillary rise behavior of the NCF 0° can be given similarly to the NCF +/- 45° material. The binder covers the whole area between the layers. The fluid is forced to rise within the rovings and does not need to fill pores between the layers. Hence, the rise velocity increases after activation.

Looking at the weaves and their capillary behaviors, one must also consider the fiber areal weight, resp. the fiber volume content of the whole stack. While for the satin weave these values are high, the twill weave with low FAW shows large gaps (see Fig. 4). Since the binder is only acting locally for both materials, the presence and the distribution of pores have to be taken into account for the satin weave, the capillary rise velocity rises slightly after activation. This can be explained that for this densely packed stack, some of the remaining pores are filled with binder. They do not need to be filled by the fluid which then can rise faster. The twill weave however shows a slower rise after activation. Here, the pores are still relatively large even when compacted. It can be assumed that the binder blocking capillary tubes within the rovings has a larger effect on the overall flow behavior.

Fig. 5
figure 5

Schematic of binder influence on capillary flow channels in the tested fabrics

Full size image

Figure 5 shows on the left side a schematic of the binder distribution between two rovings. The binder particles block space between adjacent rovings. During the capillary rise experiment, the empty space between the particles has to be filled by peripheral flow through the neighboring capillary tubes. Depending on temperature, the binder’s viscosity decreases. Liquid binder fills some of the pores between the rovings and gets absorbed by them with elevating temperature.

It can also be assumed that the distance between the layers decreases due to the acting pressure and the binder’s low viscosity. A uniform layer of binder will be established between the fabrics which leads to less pore space and hence a faster and more directed flow of the fluid. The pictures in Fig. 5 show the cross-section of a single +/- 45° layer activated at 140 °C. At the edges, the rovings appear to be more densely packed. Further magnification shows additional (binder) material in these areas. These images support the assumption of the creation of a thin interlayer binder film at high activation temperatures. This means that the capillary flow is directed within one single layer of the fabric. Also, porosities between the layers or the rovings of a single layer are already filled with binder. The test liquid is forced to rise within the capillary tubes under a 45° angle.

As discussed in [23], the peripheral flow reduces the capillary rise velocity. Specimens activated between 80 and 120 °C still allow this. Additionally, the molten binder partially blocks the pores between the layers and hence hinders the capillary flow.

In Fig. 2a) the comparison of the curves to the expected value according to the Lucas-Washburn equation is shown. It is in good accordance with the curve at room temperature but is significantly different from the curves without an activated binder which could block or divert the fluid flow.

The left side of Fig. 5 shows schematics of the binder influence on the tested weaves. As discussed before the binder particles are larger and not able to create a continuous film. Depending on the fabric’s and the resulting stack’s structure and FAW/FVC, the binder can fill pores or block capillary tubes within the roving. Both these effects influence the capillary flow contradictory and can either lead to effects like fingering or avoid them. Influences like the fabric type, the FAW, the compaction pressure, binder distribution, and binder particle size determine the capillary flow which makes it difficult to describe a general effect of binder presence/activation.

4 Conclusion

Capillary rise experiments of bindered carbon fiber fabrics showed inconclusive results. For specimens activated in a vacuum bag on a heating plate, further investigation is necessary: Similar to the studies on permeability, various effects influence the capillary flow, such as blocking of flow channels and filling of pores which are difficult to separate. An extensive study on the binder flow mechanism in preforms should be performed investigating permeability and capillary flow under the influence of temperature, compaction, and fabric structure.

Considering the specimen activated and compressed in an autoclave a deceleration of capillary flow could be noticed. The development of an interlayer film blocking peripheral flow has a larger influence than the blocking of some capillary tubes in the edge area of the rovings by binder material. Here, it will be interesting to see if this behavior can be maintained for other geometries than a flat plate, where a separating film cannot easily be established (e.g. sharp edges, curvature, bulk regions).