Two-stage wicking of yarns at the fiber scale investigated by synchrotron X-ray phase-contrast fast tomography

Abstract

With synchrotron X-ray phase-contrast computed tomography, we document the wicking process at the fiber scale in cotton, polyethylene terephthalate and polypropylene yarns. A new segmentation procedure is developed, allowing a clear separation of the water and the fiber in the reconstructed images. From the water configurations, we obtain moisture content profiles over the height of the yarn and time-resolved three-dimensional visualization of the wicking process. The water filling over the height of the yarn is highly non-uniform, since the available pore space varies strongly along the yarn due to the twisting of the yarn. For the first time, a wicking in two stages is observed: an initial fast unsaturated wetting along the fiber direction followed by a main saturated flow characterized by large jumps in moisture content at discrete time steps. These jumps occur when large pore segments become filled suddenly from multiple entry points through small size throats connecting different pore segments.

Keywords

fiber yarn materials measurement X-ray tomography wetting wicking

Yarns are the building blocks of textiles and can present a large variety of composition and configuration. These bundles of fibers, either twisted or spun together, can count from a few to hundreds of fibers. Yarns can be made of natural (cotton, wool) or synthetic materials (polymers such as polyolefins, polyesters, acrylic, etc). When these fibers show a high wettability, liquid will be taken up by capillary forces when a meniscus is formed in the pore space between the fibers. This spontaneous liquid imbibition in textiles is called wicking.¹ Given its practical relevance, wicking in yarns has been studied for a long time. The poor management of wicking can lead to serious injuries, such as steam burns in firefighters,^2,3 blister formation for long distance hiking^4,5 and decubitus in immobile patients.⁶ By 1959, an extensive study had been carried out compared liquid wicking in typical yarns, studying the effect of liquid viscosity and surface tension, contact angle and the geometry of both yarn and yarn/liquid systems. The authors provided wicking rates in terms of distance travel per minute for the liquid front.⁷ An extensive review, covering the work from the 1950s to the 1990s, classified the different wicking processes in yarn: with and without absorption; in the presence of an infinite or limited reservoir; along or across the yarn. Of interest is the observation that the liquid front in nonhomogeneous capillary systems, such as fibrous assemblies, was seen to advance by jumps, which are attributed to the irregular capillary spaces within the yarn.⁸ From this point, several studies focused on the analysis of the specific liquid distribution in the yarn pore space during wicking. Using a liquid (polyethylene glycol) colored at various dye concentrations, photographs of the yarn were interpreted, where intensity values were used as a measure of liquid saturation. The influence of yarn torsion, which determines the yarn pore size, on liquid distribution in the yarns has been determined.⁹ Torsion not only affects the pore size but also the length of the capillary path, as shown with open-packed yarns made of three layers of regular fibers.¹⁰ Despite the known role of heterogeneity in pore size, shape and orientation on the wicking process in yarn, even recent works only report the tracking of the water front; the actual water configuration within the yarn during unsaturated flow has yet to be documented (e.g. Nyoni and Brook,¹¹ Wang et al.,¹² Jad et al.¹³ and Das et al.¹⁴). We note that wicking in microsystems other than yarns is also mainly reported in terms of the displacement of liquid front, for example for microfluid waffles,¹⁵ nanofiber web¹⁶ and swelling flax and viscose fibers, where swelling is documented by environmental scanning electron microscopy.¹⁷

X-ray computed tomography (CT) is increasingly used to more accurately capture the configurations of liquid displacement inside a porous material. By using a synchrotron source for the X-rays, the acquisition time of a full tomogram can be reduced to under a second with a spatial resolution ranging from 2 to 20 micrometers (e.g. the sub-second time-resolved visualization of foam¹⁸). Other works imaged salt solution flow in a sandstone at a temporal resolution of 1 and 16.8 s,^19,20 and water visualization in a gas diffusion layer of a fuel cell at an 11 s temporal resolution.²¹ Further, the beam coherence of synchrotron X-rays allows the use of phase-contrast techniques to enhance the signal of low absorbing materials, such as water and many polymer materials. Three-dimensional (3D) imaging provides information about the configuration of the liquid within the porous medium and, with the recent fast tomography developments, variation with time can adequately be captured. We mention that laboratory-based systems have also evolved to provide CT at the sub-minute time scale (with the application of liquid in sandstones²²), but they do not allow the time resolution achieved in this study. One aspect of synchrotron X-ray CT (and, in reality, all imaging systems) is the trade-off between space resolution and field of view (FOV) size. While synchrotron X-ray CT is adequate regarding space resolution and FOV size for yarns, it would have too small a FOV to document wicking in fabrics. Therefore, for fabrics other methods are more appropriate, for example neutron radiography.^23,24

The measurement of wicking in yarns has relied mostly on visual methods (e.g. Taheri et al.²⁵), usually with the aid of colorants or electrical contact methods (e.g. Ansari and Kish²⁶); see also the review paper by Parada et al.²⁷ Such methods give only a one-dimensional view of wicking. As described above, X-ray tomographic imaging allows the full tridimensional (3D) imaging of a sample and, with the low acquisition time and phase-contrast enhancement provided by synchrotron X-ray CT, it is possible to image the dynamic wicking process in a yarn. We present here the application of this technique to wicking in yarns and provide the first full 3D time-resolved documentation of the liquid and fiber configurations during wicking in different yarns. The objectives of this work are to document in a time-resolved way the configurations of water in yarns during wicking and to understand the effect of the complex yarn geometry on these evolving water configurations.

In this paper, first we describe the materials used as well as the previous preparation of the materials. Next, we present details of the synchrotron X-ray phase-contrast microscopic tomography technique and the imaging parameters used for this experiment. We also present our setup and imaging procedure. Then, the image processing and analysis are presented, showing how the water and the fibers are segmented. The result section follows, where we present the dynamics of moisture content evolution for all samples and document in detail the wetting dynamics in a yarn segment. Finally, the conclusion highlights the new developments and findings performed in this work as well as possible further research.

Materials and methods

Materials

Yarns of polyethylene terephthalate (PET), polypropylene (PP) and cotton at different twisting levels were investigated. Table 1 summarizes the yarn properties. The samples were all washed and rinsed multiple times to remove any lubricant or chemical residue from manufacturing. The contact angles were measured by holding the sample perpendicular in a reservoir and measuring the angles on both sides of the yarn at the water surface (see Figure 1). We used the tangent method and performed the measurements with the Fiji distribution of ImageJ.²⁸ At least three images of each yarn were captured and the values reported were averaged for the different twisting levels, since no significant effect of twisting level on the contact angle was observed. The equilibrium contact angle ranges between 21 degrees for cotton and around 50 degrees for PET and PP.
Figure 1.
Contact angle measurement examples for each material. PET: polyethylene terephthalate; PP: polypropylene.

Table 1.
Yarn material properties

Material Cross-section Yarn linear density (dtex or g/10 km) Number of fibers (filaments) in yarn Single fiber diameter (density g/cm³) Twisting level (tpm) Contact angle (°)

Cotton – 148 (Ne 40) NA – 1000 21 ± 7

PET Circular 167 32 22 µm (1.35) 100, 200, 300 53 ± 16

PP Hollow circular 110 33 22 µm (0.85) 110, 230 50 ± 14

Ne: a cotton count number (English number) equivalent to the number of hanks (770 m) in one pound of yarn; tpm: number of twists per meter length of yarn; PET: polyethylene terephthalate; PP: polypropylene.

Synchrotron X-ray phase-contrast fast tomography

Synchrotron X-ray tomography is a non-destructive 3D imaging technique that uses a synchrotron accelerator as the source of the penetrating beam. The high brilliance of the synchrotron light source allows low acquisition times (i.e. fast frame rate) and, at the same time, the high coherence allows for phase-contrast techniques. Both characteristics are important for our experiment, since the wicking process is fast (of the scale of seconds for a 5 mm sample length) and the materials involved absorb X-rays poorly, making phase-contrast techniques necessary to distinguish textile material, air and liquid.

The experiment was performed at the TOMCAT beamline of the Swiss Light Source (SLS) of the Paul Scherrer Institute (PSI) in Villigen, Switzerland. The beamline light was provided by a 2.9 T superbend at a critical energy of 11.1 keV, which then passed through a fixed-exit double crystal multilayer monochromator (DCMM) that transmitted the beam at an energy between 8 and 45 keV. For our experiment, the energy was set at 14 keV. After interacting with the sample, the beam reached a 150 µm thick Ce-doped Lu₃Al₅O₁₂ scintillator placed at a distance of 270 mm downstream of the sample, which converted the X-ray intensity into visible light. This visible light image was magnified by a high numerical aperture optical light microscope (Elya Solutions) with a magnification of 3.78× and then captured by a 2016 × 2016 pixel CMOS imaging chip with a pixel size of 11 um, resulting in an effective pixel size of 2.91 µm. The imaging chip was part of the GigaFRoST camera,²⁹ which allowed fast read-out times and data streams of up to 8 GB/s. Since our samples had a high aspect ratio, the images were promptly cropped to 239 × 1647 pixels (large pixel count in the yarn length direction) to reduce file size and reconstruction time later on, yielding a FOV of 0.7 × 4.8 mm² (h × v).

The sample was fixed inside a sample holder consisting of a Kapton tube and a reservoir (Figure 2). Kapton was selected since it has low absorption of X-rays. The tube length was 32 mm and the diameter was 5 mm. The yarn sample with a diameter of less than 300 µm was suspended in the middle of the tube with a pre-tension of 80 mN. The pre-tension was applied with a mass (8.13 g) at the end of the yarn during fixation. The sample was fixed by a pair of screws in the top and bottom part of the sample holder. During the experiment, it was observed that the yarn sample in the holder remained immobile and could be considered as fixed. The reservoir was made of aluminum and contained 2.9 ml of liquid. At the base of the reservoir, four holes in the holder allowed the water to reach the sample. The entire sample holder was held in place by friction, that is, no type of glue or other adhesive was used to connect the Kapton tube to the reservoir or the top cap. The samples already in the sample holders were kept at 50℃ before the experiment to ensure a dry state.
Figure 2.
Photograph of the sample holder at the stage (a) and schematic of the experimental setup (b). Water is delivered to the reservoir by a remote-controlled syringe pump (not depicted) connected to a needle. The schematic (not to scale) is shown as a side cut view (top) and top cut view (bottom). FOV: field of view.

Prior to the experiment, the complete sample holder with reservoir was attached to a rotating table. Firstly, 50 dark field images (no beam) and 100 flat field images (with beam but no sample) were taken to correct for non-uniformity of the beam and different pixel sensitivities of the CMOS camera. Secondly, the sample was placed at the FOV and was accelerated until the rotation speed is 0.83 Hz (i.e. rotation period of 1.2 s). Thirdly, the acquisition and the water delivery were started simultaneously. The remote-controlled syringe pump was controlled by a manual trigger that could be activated from outside the experimental hutch. The needle was set to deliver 2 ml of de-ionized water to the reservoir in 4 s and did not touch the reservoir. We estimated the difference in height between the edge and the center of the water surface due to rotation to be less than 0.08 mm, and thus a quasi-flat water surface was present in the reservoir. Finally, the imaging process started when the water reached the bottom of the FOV, which was positioned at 17 mm above the water level. Images were recorded continuously until the sample in the FOV was fully wet. The exposure time was set to 1 ms and the number of projections (per 180°) was 600. One-hundred-and-eighty degrees of each rotation were used for imaging, resulting in a time resolution of 1.2 s for one scan. The total acquisition time varied with the sample but amounted to 11 min maximally.

Temperature and relative humidity were monitored at three different locations inside the experimental room and the values recorded on average were 26.3 ± 0.3℃ and 23 ± 1%RH.

Image processing and analysis

The image processing procedure consisted of reconstruction, registration, segmentation and removal of background effects through masking and determination of water configuration, as described below. The analysis was done using Python 2.7 and the Fiji distribution of ImageJ,²⁸ unless otherwise mentioned.

The radiographs were first reconstructed using the Paganin algorithm, which applies a Fourier approach to solving the transport of intensity equation for near-field free propagation to obtain the thickness, phase and absorption from the projected images.³⁰ Then, since some samples slightly moved or rotated during the experiment, we corrected the images by applying a rigid-body registration per slice (along the yarn length) using the openCV library.³¹

The resulting 8-bit 3D images were then segmented to separate the water and the fiber. Due to the complex water configurations during wicking, we found that standard segmentation did not give satisfactory results, mainly due to the high noise ratio. Therefore, we developed a new method for segmentation of the images, specific for wicking in yarns. The method is based on the analysis of the time series of the gray-level referred to as intensity for each voxel. Table 2 and Figure 3 show the procedure for identifying the jump. Figure 3 shows, as an example, the original time signal of the recorded intensity for one voxel (black line). We observed that the intensity signal was quite noisy, showing both higher and lower frequency variations as well as a clear jump in intensity at 200 s. The jump in intensity could be attributed to the filling of the voxel with water, indicating a jump from the dry to the wet condition. We developed a procedure to identify the time when the filling of the voxel occurs. Firstly, a low pass filter was applied to remove high-frequency noise, obtaining the red line. Next, the gradient of the filtered time series was calculated (solid blue line) and the location of the maximum gradient was determined (dashed vertical line) and was identified as a possible jump in intensity, indicating the filling of the voxel with water. Then, the average intensity in the previous 20 time steps and the following 20 time steps was calculated (Figure 3, green line). This range was manually specified as a good value based on the variations of intensity from the original images. If the difference between the average intensities before and after the jump was larger than a specific threshold, the jump was considered to be a real filling of the voxel at the time of the maximum gradient. We found for our specific case of wicking of yarns a threshold value of 50 to be adequate to identify these jumps, since too low a value would select too many jumps, which cannot be attributed to the filling of a voxel, while too high a value would not allow one to find a jump. If a real jump in intensity was recognized, we assumed that the voxel was totally filled with water at this time step and that, for all time steps before the jump, the voxel could be considered to be empty, while for all time steps following the jump the voxel was considered to be filled with water. In the case that the jumps over the whole time period were too small, we considered the voxel to be part of the pore space that remained empty. In the case that the mean intensity remained always above 100, we assumed that this voxel belonged to the fiber region. We remark that this procedure is quite specific for the particular problem of wicking in yarns, where the pore space becomes filled, showing jumps in moisture content (as shown below). Such a method would not be applicable for a wetting process where voxels are gradually filled, as is the case of capillary uptake in common porous materials showing a slow moving moisture front with a whole range of moisture contents.
Figure 3.
Sample intensity value versus time for one pixel and demonstration of the steps toward finding a transition from dry to wet: original signal (black); filtered signal (red); gradient of filtered signal (solid blue); position of the peak gradient (dashed blue); values before and after transition (green). (Color online only.)

Table 2.
Algorithm for water and fiber segmentation

Algorithm Water and fiber segmentation

1: for each voxel in 3D image do

2: Apply low pass filter to time series

3: Calculate gradient of filtered time series

4: Find maximum gradient position

5: Calculate average intensity in 20 points before and after maximum gradient position

6: if jump height >50 then

7: Assign all points before jump as 0 and all points after jump as 1

8: else

9: if mean intensity in whole time series >100 then

10: Assign all points in time series to 1 (fiber)

11: else

12: Assign all points in time series to 0 (air)

13: end if

14: end if

15: end for

To remove background information from the images for further post-processing and easy determination of the moisture content, the last image of the fully wet sample was used to create a mask of the images. The mask was obtained by first binarizing the image using a threshold value. Then, using closing morphological operations, the gaps between the fibers were closed until all holes in the inside of the yarn were removed, obtaining a mask image. This mask was then multiplied to all segmented images to remove any effect of the background. Finally, in the segmented initial dry image, all voxels identified as ‘full’ represented voxels filled by fibers. Therefore, this image was used as a reference and was subtracted from all subsequent images. The resulting images consisted only of voxels identified as water and as yarn. Figure 4 depicts a time series of three slices, taken at the bottom, middle and top of an imaged yarn, before and after the segmentation procedure.
Figure 4.
Time evolution of moisture at three different heights for the polyethylene terephthalate 300 tpm sample. The top three rows show cross-section slices of the original images; the bottom three rows show the segmented images. In the segmented images the yarn is shown in white and water is shown in blue. The height shown is relative to the field of view (FOV) and the time shown is relative to the first time step where water is identified in the FOV. (Color online only.)

The images were then further analyzed to obtain the temporal variation of moisture content along the sample height using the procedure illustrated in Figure 5. This figure shows the FOV of a measured sample (PET 300 tpm) as an example. Figure 5(a) left shows the segmented image of the fibers in the dry state before the start of the wicking process. In this figure, all the voxels shown are identified as filled by fiber. This image was used as a reference image (or ‘dry’ image). Figure 5(a) middle shows a segmented image at time t during the wicking process, showing both fiber and water-filled voxels. The water-filled pixels were identified using the segmentation procedure explained above. Due to correction for rigid-body registration per slice, we could then subtract the segmented reference image from the segmented image at time t, obtaining as such only the water-filled regions (Figure 5(a) right). The resulting images show the water configuration, that is, the voxels that are totally filled by water at time t. This work is here visualized with Avizo Fire.
Figure 5.
(a) Three-dimensional view of a selected small region of 0.7 × 0.7 × 1.8 mm³: left, ‘dry’ image showing the fibers; middle, ‘wet’ image at time t during the wicking process with fibers in gray and water in blue; right, subtracted image (wet – dry) showing only the voxels filled with water. (b) Profiles of moisture content per unit length versus height for the polyethylene terephthalate (PET) 300 tpm sample for the full field of view (FOV), where the colors show different time steps during the wicking process. (c) Total moisture content in the full FOV versus time for the PET 300 tpm sample. (d) Time evolution of average moisture content in five consecutive sections (0: top, 4: bottom) along the height of the sample. Note that the colors in (b) and (d) are unrelated. (Color online only.)

Based on these images containing only water, we could determine the moisture content per unit length over the height, by summing up all the water-filled voxel volumes at a certain height and converting these values to moisture content by multiplying with water density. As an example, Figure 5(b) shows the moisture content profiles over the height of the FOV versus time where the colors ranging from blue to red represent different time steps during the wicking process. The lower part of the plot shows moisture content closer to the bottom of the FOV, while the upper part shows a region close to the top of the FOV. Blue colors are earlier time steps, while red colors are later time steps. We could also determine the time evolution of the total moisture amount in the FOV or in sections of the FOV by summing all the voxels identified as water and converting to water mass by multiplying the total voxel volume with the bulk water density. As an example, Figure 5(c) presents the moisture content evolution over time for the total FOV, while Figure 5(d) gives the wicking process in five consecutive sections along the height of the FOV.

The measured moisture contents reported in this paper are a lower limit to the total amount of water detected. In addition to this amount, there was a certain amount of undetected water in the system, which was below the dataset resolution (the pixel size is 2.91 µm). In order to get an estimate of the amount of undetected water, we can consider a very dense packing of the fibers, meaning the cross-section of the yarn to be represented by a two-dimensional high dense packing of circles of diameter D. Such a highly dense structure yields a density of 0.9 and a porosity of 10%, resulting in a spacing between circles that can fit a smaller circle of diameter 0.15D. For our specific geometry, our fibers have an approximate diameter D of 22 µm diameter, resulting in pores of a minimum 3.3 µm diameter. This value is close to our resolution (2.91 µm) and thus is difficult to detect. In this worst case scenario of 10% porosity and high packing, the maximum undetected water amount is 10%. Our real geometry, however, has a much higher pore space in between fibers than a dense packing, so 10% undetectable water amount is an overestimation. Regardless of such possibility of the undetected amount, the segmentation method used in this paper provides a reliable way of detecting changes in intensity only due to water. Efforts in providing a direct measuring of water uptake in yarns have failed due to the small quantity of water involved in such experiments.

Results

The time evolution of the total moisture content for all samples is presented in Figure 6. Note that zero time is the time that the water is first present in the FOV and that the FOV is positioned 17 mm above the water level. Cotton shows a quite fast uptake indicated by a single steep jump followed by a more gradual uptake, with a total uptake volume of 5.5 e-5 g in the FOV or 11 mg/m of yarn length. The uptake behavior of PET shows more gradual uptake curves with many small consecutive jumps (100-1b, 100-2, 300), or steep jumps followed by a more gradual uptake (100-1a, 200), with a total amount of water taken up between 1 e-5 and 4 e-5 g in the FOV or 2–8 mg/m of yarn length. Also for PP, the uptake can be fast (110-1, 230) or more gradual (110-2), with a total mass ranging from 1 e-5 to 4 e-5 g in the FOV or 2–8 mg/m of yarn length.
Figure 6.
Total moisture content versus time in the full field of view (FOV). The numbers and letters in brackets represent repeats with same sample [letters] and different samples from the same yarn [numbers], that is, ‘PET, 100 tpm, [1a]’ and ‘PET, 100 tpm, [1b]’ are the exact same sample tested twice, while ‘PP, 110 tpm, [1]’ and ‘PP, 110 tpm, [2]’ are two different samples from the same yarn. PET: polyethylene terephthalate; PP: polypropylene.

Taking a closer look on PET samples of the same twisting level of 100 tpm (Figure 7), we see the same different behavior even for samples from the same bobbin. The maximum moisture content at the end of the wicking process and the time to reach this maximum is quite different, as well as the average uptake rate indicated by the slope of the fitted linear curve. Even repeats of the same exact sample (• and ▪ samples) show different behavior. This means that the wicking of yarns may differ strongly from sample to sample and from repeat to repeat. Differences between different repeats may be due to changes in fiber geometry due to sample manipulation between different repeats or due to impurities introduced along the fiber surface affecting wettability after wetting and drying of the samples. However, in general, we may conclude that the wicking process proceeds in different stages, where one stage is characterized by large jumps in moisture content, while the other stage shows a more gradual increase of moisture content. Therefore, we analyzed in more detail the moisture content (per unit length) profiles over the height of the FOV for different times (Figure 8). The colors ranging from blue to red represent the different times during the wicking process. For all samples, we observe that the moisture content varies considerably over the height of the FOV, showing irregular profiles. In particular, the final moisture content distribution at the end of the wicking process shows considerable variation in moisture content showing single or multiple peaks over the height of the sample. This indicates that the pore space filled by water at the end of the wicking process varies from position to position over the height of the yarn. This variation in filled pore space may be attributed to differences in pore space available between the fibers, possibly due to the twist of the yarns.
Figure 7.
Total moisture content in full field of view (FOV) for polyethylene terephthalate (PET) yarns with 100 tpm. Lines represent linear fit and the blue region represents the standard deviation of the linear fit. (Color online only.)

Figure 8.
Moisture content profiles over the height of the field of view for all samples. Time is represented by color. Circled regions point to sudden jumps in moisture content. Numbers and letters in brackets represent repeats [letters] and multiple samples from the same yarn [numbers]. PET: polyethylene terephthalate; PP: polypropylene. (Color online only.)

It is important to note that at early times the moisture content profiles (blue lines in Figure 8) may already range over a considerable height of the FOV, showing that some water paths may already develop at an early time, reaching over the total height of the FOV. This is in contrast with the uptake process in common porous materials, where a moisture front gradually moves up over time along the uptake direction. Here we observe a preferential wetting process that reaches the top of the FOV well before the maximum moisture content is reached. This first stage wetting process is followed by a second stage, where the moisture content further increases with time over all heights. When we compare the moisture content distributions over height for different times, we observe that the changes in moisture content between different times are not uniform over time or over the height of the specimen. This means that the rates of wicking at different heights are different and not constant over time. In particular, we note some sudden increases in moisture content at certain heights and time steps, as indicated by the dashed circles in Figure 8, indicating large changes in moisture content happening over short times. The large jumps in moisture content in the global wicking curves, as observed in Figure 6, may be attributed to these sudden changes in moisture content along the height of the yarn.

We performed next a more detailed analysis of the moisture wicking process for a particular sample, PET 300 tpm, which allowed us to analyze in more detail the local jumps in moisture content. For this sample, Figures 5(c) and (d) present the time evolution of the moisture content in the total FOV and in five consecutive sections along the height of the FOV, respectively. Figure 5(b) shows the evolution of the moisture content profiles over time over the height of the FOV. In addition, Figure 9 gives snapshots of the water-filled configurations at different times starting from t = 72 s. Figures 5(d) and 9 show that the wicking process starts at the bottom of the FOV followed by a fast increase in moisture content in the bottom section 4 (at t = 50 s). This wetting of the first section is quickly followed by water penetration in all the other sections 0–3. Figure 9 shows that after 96 s the liquid water already reaches the top of the FOV. Note that the figure shows some disconnected water regions. This indicates that there are connecting water paths that are below our resolution of water recognition. Figure 5(d) shows that between around t = 130 and 170 s the wicking process almost stops in all sections, followed by a sudden large jump in moisture content in all sections at t = 174 s. It is remarkable that this jump in moisture content occurs almost at the same time in all sections over the height of the FOV, indicating a sudden penetration of liquid along a path in the yarn direction. Note also that the highest jump in moisture content occurs in sections 0–2 at the top of the FOV.
Figure 9.
Water configurations versus time of the polyethylene terephthalate 300 tpm sample at different time steps in the full field of view. Dashed lines represent the region in time and space where a major jump in moisture content occurs.

To study this sudden penetration of liquid in more detail, we selected the region marked by dashed lines in Figure 9, where a high jump in moisture content occurs. The region is located at the top of the FOV and is 1.8 mm high (630 slices in the region). We segmented the pore space in this region into several pore regions using the Thickness Map algorithm in Aviso, which defines thickness as the diameter of the largest enclosed ball that contains the pixel.³² We used the final filling geometry to perform a segmentation of the pore space in different pore regions. Out of these pore regions, we selected four main pore segments that were filled during the jump and analyzed the water filling process versus time between time steps 174 and 198 s. Figure 10 shows the water filling process of the four main pore regions indicated with different colors at six different time steps. The jump in moisture content in the red pore region shows first a water filling extending in the fiber direction over 12 s (left three images). Then, the water filling in the red pore shows a further increase (thickening) of the water-filled zones over 12 s (right three images). The same stages in water filling can be observed in the blue pore region: first an extension of the water filling in the fiber direction followed by a thickening process. We observe that the filling of a pore region may occur from more than one entry point, and can even happen in a direction opposite to the yarn direction. The fast filling process of several pore regions at the same time from multiple entries corresponds to the observed jump in moisture content in Figure 5(c). As an example, Figure 11 visualizes the filling of two single pore segments, showing that these pore regions fill mainly in a few time steps around the moisture content jump. Detailed analysis further shows that the pore space is built up from long elongated large pore regions along the yarn direction, which are connected by smaller pores (throats), where probably the smallest are under our resolution. This would mean that the jumps in the wicking process are controlled by small size throats and that the main pore regions are only invaded when the resistance of the throats is overcome.
Figure 10.
Three-dimensional view of water configurations in the top part of the sample, as marked in Figure 9. Four main water regions are depicted at six time steps (from left to right) of 174, 179, 183, 188, 193 and 198 s. Each color represents one connected water region. The vertical dimension is compressed to facilitate visualization. (Color online only.)

Figure 11.
Detailed view of two connected water regions from Figure 10 at the six different time steps (from left to right) of 174, 179, 183, 188, 193. The vertical dimension is compressed to facilitate visualization.

Conclusions

In this paper, we demonstrate the use of synchrotron-based phase-contrast X-ray tomography for the imaging of the wicking process in yarns consisting of different twisted fibers. A new imaging processing technique for the segmentation of the water phase is presented. This new procedure, especially developed for the wicking of yarns, is based on the detection of jumps in intensity to segment liquid and fiber, allowing one to identify when a voxel is filled with water. The resulting time-resolved 3D images of water configurations allow us to analyze the wicking process of yarns at the fiber scale. To the best of our knowledge, we first report in this study the observation of two stages in the wicking process. Firstly, a fast flow occurs propagating quickly along the fiber directing, reaching the top of the sample much quicker than the time needed for the complete filling of the available pore space. We believe this first stage in the wicking process starts with an unsaturated flow (film/corner flow) around the fibers partly under our resolution of detection. In the second stage of wicking, large pore regions become filled by water and become connected to other major pore regions. This second stage process is not only characterized by a vertical flow process along the yarn direction, as in the case in the first stage wetting process, but is also characterized by sudden invasions of major elongated pore regions through small throats. This invasion process occurs from multiple entry points and is controlled by the size of the connecting throats, allowing main pore regions to be only invaded when the resistance of the throats is overcome. This process of sudden filling of large pores explains the sudden jumps in moisture content both along the yarn length and in time. The appearance of these small size throats in between large elongated pore size regions along the fiber direction may be attributed to the yarn torsion during the manufacturing process. An important observation is the large heterogeneity in wicking between the different yarn samples and even between repeats of the same sample, meaning wicking in yarns is probably controlled by small changes in the pore space controlled by the torsion exerted on the yarns. Another possible source of the heterogeneity could be the presence of impurities on the fiber surface affecting wettability. Case in point, PP is usually hydrophobic, but in our PP samples we observed a contact angle <90° and the PP yarns did wick enough water to be able to measure it.

From all possible sample parameters, we only identify the material type as having an impact on wicking behavior. This does not, necessarily, mean that other parameters (such as twisting level) do not have any impact. Further research is needed to unravel the impact of other parameters, testing a larger number of samples to obtain better statistics regarding the influence of twisting level including a wider range of twisting levels. Other parameters to be investigated are the number and size of fibers and different contact angles, either by varying the liquid or by coating the samples before measurement. Finally, an effort to simulate the wicking behavior using the measured yarn geometry seems worthwhile, using approaches such as pore networks or Lattice Boltzmann methods.

Material	Cross-section	Yarn linear density (dtex or g/10 km)	Number of fibers (filaments) in yarn	Single fiber diameter (density g/cm³)	Twisting level (tpm)	Contact angle (°)
Cotton	–	148 (Ne 40)	NA	–	1000	21 ± 7
PET	Circular	167	32	22 µm (1.35)	100, 200, 300	53 ± 16
PP	Hollow circular	110	33	22 µm (0.85)	110, 230	50 ± 14

Algorithm	Water and fiber segmentation
1:	for each voxel in 3D image do
2:	Apply low pass filter to time series
3:	Calculate gradient of filtered time series
4:	Find maximum gradient position
5:	Calculate average intensity in 20 points before and after maximum gradient position
6:	if jump height >50 then
7:	Assign all points before jump as 0 and all points after jump as 1
8:	else
9:	if mean intensity in whole time series >100 then
10:	Assign all points in time series to 1 (fiber)
11:	else
12:	Assign all points in time series to 0 (air)
13:	end if
14:	end if
15:	end for

Footnotes

Acknowledgments

The authors would like to thank Bäumlin & Ernst AG, Chemosvit Fibrochem and LEGS for providing the yarns. We thank Prof Artan Sinoimeri, from LPMT at Université de Haute-Alsace (Mulhouse, France), for helping us with the re-twisting of cotton yarns. We also thank Stefan Carl, from Empa, for the support in developing the different experimental devices and Rajmund Mokso, formerly from PSI, for support during tomography experiments. We acknowledge the Paul Scherrer Institut, Villigen, Switzerland for provision of synchrotron radiation beamtime at the TOMCAT beamline X02DA of the SLS.

Declaration of conflicting interests

The authors declared no potential conflict of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the ETHZürich (Grant # 0-20909-13).

References

Masoodi

Pillai

. Evaporation and wicking. In: Masoodi R and Pillai KM (eds) Wicking in porous materials : traditional and modern modeling approaches, Boca Raton: CRC Press, 2013, pp. 201–236.

Keiser

Wyss

Rossi

. Analysis of steam formation and migration in firefighters’ protective clothing using X-ray radiography. Int J Occup Saf Ergon 2010; 16: 217–229.

Zhang H, Song G, Gu Y, et al. Effect of moisture content on thermal protective performance of fabric assemblies by a stored energy approach under flash exposure. Text Res J 2018; 88: 1847–1861.

Baussan

Bueno

Rossi

, et al. Analysis of current running sock structures with regard to blister prevention. Text Res J 2012; 83: 836–848.

Bogerd

Niedermann

Brühwiler

, et al. The effect of two sock fabrics on perception and physiological parameters associated with blister incidence: a field study. Ann Occup Hyg 2012; 56: 481–488.

Derler

Rao

Ballistreri

, et al. Medical textiles with low friction for decubitus prevention. Tribol Int 2012; 46: 208–214.

Minor

Schwartz

Wulkow

, et al. The migration of liquids in textile assemblies: Part II: the wicking of liquids in yams. Text Res J 1959; 29: 931–939.

Kissa

. Wetting and wicking. Text Res J 1996; 66: 660–668.

Perwuelz

Casetta

Caze

. Liquid organisation during capillary rise in yarns—influence of yarn torsion. Polym Test 2001; 20: 553–561.

10.

Liu

Choi

. Wicking in twisted yarns. J Colloid Interface Sci 2008; 318: 134–139.

11.

Nyoni

Brook

. Wicking mechanisms in yarns—the key to fabric wicking performance. J Text Inst 2006; 97: 119–128.

12.

Wang

Zha

Wang

. Study on the wicking property of polyester filament yarns. Fibers Polym 2008; 9: 97–100.

13.

Jad

MSM

Ravandi

SAH

Tavanai

, et al. Wicking phenomenon in polyacrylonitrile nanofiber yarn. Fibers Polym 2011; 12: 801–807.

14.

Das

Kothari

, et al. Development of mathematical model to predict vertical wicking behaviour. Part I: flow through yarn. J Text Inst 2011; 102: 957–970.

15.

Chen

Melvin

Weislogel

, et al. Design, fabrication, and testing of microporous wicking structure. Microelectron Eng 2008; 85: 1027–1030.

16.

Callegari

Tyomkin

Kornev

, et al. Absorption and transport properties of ultra-fine cellulose webs. J Colloid Interface Sci 2011; 353: 290–293.

17.

Stuart

McCall

Sharma

HSS

, et al. Modelling of wicking and moisture interactions of flax and viscose fibres. Carbohydr Polym 2015; 123: 359–368.

18.

Mokso

Marone

Haberthür

, et al. Following dynamic processes by X-ray tomographic microscopy with sub-second temporal resolution. AIP Conf Proc 2010; 1365: 38–41.

19.

Youssef

Deschamps

Dautriat

, et al. 4D imaging of fluid flow dynamics in natural porous media with ultra-fast X-ray microtomography. Int Symp Core Anal 2013, pp. 1–12.

20.

Berg

Ott

Klapp

, et al. Real-time 3D imaging of Haines jumps in porous media flow. Proc Natl Acad Sci 2013; 110: 3755–3759.

21.

Eller

Roth

Marone

, et al. Implications of polymer electrolyte fuel cell exposure to synchrotron radiation on gas diffusion layer water distribution. J Power Sources 2014; 245: 796–800.

22.

Bultreys

Boone

, et al. Real-time visualization of Haines jumps in sandstone with laboratory-based microcomputed tomography. Water Resour Res 2015; 51: 8668–8676.

23.

Parada

Vontobel

Rossi

, et al. Dynamic wicking process in textiles. Transp Porous Media 2017; 119: 611–632.

24.

Parada

Zhou

Derome

, et al. Modeling wicking in textiles using the dual porosity approach. Text Res J 2018, pp. 1–10.

25.

Taheri

Vadood

Johari

. Investigating the effect of yarn count and twist factor on the packing density and wicking height of lyocell ring-spun yarns. Fibers Polym 2013; 14: 1548–1555.

26.

Ansari

Kish

. The wicking of water in yarn as measured by an electrical resistance technique. J Text Inst 2000; 91: 410–419.

27.

Parada

Derome

Rossi

, et al. A review on advanced imaging technologies for the quantification of wicking in textiles. Text Res J 2017; 87: 110–132.

28.

Schindelin

Arganda-Carreras

Frise

, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods 2012; 9: 676–682.

29.

Mokso

Schlepütz

Theidel

, et al. GigaFRoST: the gigabit fast readout system for tomography. J Synchrotron Radiat 2017; 24: 1250–1259.

30.

Paganin

Mayo

Gureyev

, et al. Simultaneous phase and amplitude extraction from a single defocused image of a homogeneous object. J Microsc 2002; 206: 33–40.

31.

Bradski

. The OpenCV library. Dr Dobbs J Softw Tools 2000; 25: 120–125.

32.

Hildebrand

Rüegsegger

. A new method for the model-independent assessment of thickness in three-dimensional images. J Microsc 1997; 185: 67–75.