# Designer patterned functional fibers via direct imprinting in thermal drawing

Jul 31, 2020

### Creating surface patterns on fibers with different materials

Compared to the traditional thermal fiber drawing process, the DITD process introduces a pair of rollers with desired surface structures as templates to thermally imprint surface patterns onto the drawn fiber. As sketched in Fig. 1a, the DITD process starts from feeding a preform into furnace slowly to heat it above glass transition temperature till soft. Then a neck-down region is formed in the heating zone as the softened preform is scaled down cross-sectionally and elongated into fiber under a constant downward drawing speed. A pair of patterned rollers are fixed at both sides of the fiber right below the neck-down region. Since the temperature of the neck-down region is still high enough to reshape the fiber, an exactly inverted surface pattern can be created on both sides of the drawn fiber after being directly pressed by the patterned rollers acting as templates. Meanwhile, the patterned fiber cools down rapidly because of the physical contact with the rollers, which greatly restricts subsequent distortion caused by reflow15. While the fiber is being drawn down continuously, the rollers keep rotating to imprint patterns onto fiber surfaces smoothly and repeatedly, creating a continuous surface pattern on the entire length of drawn fiber via the DITD process. A demonstration of this continuous fabrication process is shown in Supplementary Movie 1.

The surface pattern of rollers can be fabricated either by directly modifying the roller surface or simply attaching a patterned flexible substrate on the roller surface. Therefore, almost all of the processes for fabricating micro/nanostructure such as self-assembly38, lithography39,40, focused ion beam (FIB) etching41, and laser machining42,43 could be used for preparing the patterned rollers, which enables the DITD technique to fabricate fibers with a wide variety of surface patterns. In our initial experiment, a polyethylene terephthalate (PET)-based two-dimensional (2D) micro-hole array was attached on both rollers while an unpatterned polyvinylidene fluoride (PVDF) bar was chosen as the preform (see Methods for the detailed information). As a result, PVDF fiber with the inverted surface pattern, 2D microdot array, on both sides was successfully fabricated via the DITD technique, which is verified by scanning electron microscope (SEM) image in Fig. 1b. This surface pattern was further tested under illumination by white supercontinuum laser as the 2D microdot array served as a 2D diffraction grating. After the white supercontinuum laser propagated through the patterned PVDF fiber, the laser was split into several beams in the spatial domain and laser with different wavelengths was also separated in different diffraction angles in the presence of the 2D diffraction grating. Thus, a sharp 2D diffraction pattern with “rainbow” colors was formed and captured in Fig. 1c, proving the high quality of the periodic surface pattern on PVDF fiber. It is worth mentioning that the DITD technique is a high throughput process with excellent reliability to produce a single continuous patterned fiber. As exhibited in Fig. 1d, e, a 300-meter-long fiber with 2D dot array fully covered on both sides was fabricated and collected on a capstan. The diffraction pattern can be observed on the whole capstan, indicating that the surface pattern is created continuously on the entire surface of the 300-meter-long fiber.

Next, a series of fibers with various surface patterns were fabricated under the same conditions using different templates attached on the rollers. Firstly, parallel lines with both width and spacing of 5 μm in three directions (parallel, perpendicular, and 45° angle to the fiber’s axial direction) were created on the surface of PVDF fibers, as shown in Fig. 1f–h. The strict periodicity and sharp edges confirmed the high resolution of the surface pattern. Secondly, a more complex pattern (Fig. 1i) with nonuniform depth was fabricated by changing the template to a 1000-mesh stainless steel screen. Insets of Fig. 1f–i present the overview of the respective micropatterned fiber (all the scale bars in the insets are 500 μm). Further, the ability of imprinting in all directions and high resolution enable the DITD technique to produce fibers with arbitrarily designed surface pattern. As displayed in Fig. 1j, customized letters “NTU” was successfully imprinted on the fiber surface.

Knowing that almost all the thermoplastic materials are applicable to the imprinting process, the DITD process is in principle compatible with all the thermoplastic materials employed in thermal fiber drawing process, including amorphous polymers such as polycarbonate (PC), polyethyleneimine (PEI), cyclic olefin copolymer (COC), and polystyrene (PS), semicrystalline polymers such as polyether ether ketone (PEEK) and PVDF, and elastomers such as Santoprene and styrene-ethylene-butylene-styrene (SEBS). As a verification, fibers drawn from PC, PEI, PEEK, and SEBS with the same surface pattern were fabricated through the DITD process (Supplementary Fig. 6). Furthermore, the surface pattern does not influence the instinct properties of the materials. As shown in Supplementary Fig. 7, SEBS fiber with surface pattern preserved its superior flexibility as it was stretched to 600% of its original length without breaking. The high compatibility on materials selection offers the DITD technique great potentials in a wide range of applications.

### Thermal behavior study of DITD process

Studying the thermal behavior of the fiber is crucial to gain a better understanding of the DITD process. Since the thermal process of fiber drawing has been indepthly studied in earlier works15,28, we only focus on the thermal process of surface pattern imprinting after the fiber is drawn out of the neck-down region. Finite element analysis was conducted by COMSOL to simulate heat transfer among fiber, rollers, and ambient air as well as thermal radiation of all surfaces. All the geometry and parameters were defined according to the actual sizes and materials used in our experiment. Figure 2a shows the simulated temperature distribution by setting fiber temperature to 160 °C before contacting with the rollers. And the temperature of the fiber drops rapidly to ~118 °C after being pressed by rollers, as shown in the enlarged area of Fig. 2a. The rapid temperature drop during the contact is mainly caused by two factors. One factor is the thermal conductivity contrast between the roller and air. The roller we used is made of steel, whose thermal conductivity coefficient is around 30 W m−1 K−1, which is about 3 orders of magnitude higher than that of air. Thus, the heat transfer between roller and fiber is much faster than that between roller and air under the same temperature difference. The other factor is the total heat capacity contrast between roller and fiber. During the contact, the fast heat transfer leads to a rapid temperature drop of the fiber because of its low heat capacity, while the temperature of the roller only increases a little because of the high heat capacity, which helps maintain a high temperature difference. Additionally, the large roller size could also help in dissipating heat form roller to air. Thus, the temperature of the roller could be maintained at a relatively low value by absorbing heat from the fiber and dissipating heat to air simultaneously. Considering that the drawing temperature varied with different materials, we further ran a parameter sweep to obtain the temperature distribution along the fiber under different drawing temperatures. To simplify the simulation, we assumed that the temperature of fiber was constant before contacting with rollers. The temperature as a function of the distance from the neck-down region is plotted in Fig. 2b. A rapid temperature drop under all drawing temperatures is observed at the distance of 1 cm, which is the position of the rollers in our geometry model. The higher the drawing temperature used, the more the temperature drops when being pressed by rollers. Additionally, at a larger distance, the temperature keeps decreasing as the fiber leaves rollers and is drawn downward. The smaller curve slops after the fiber left the rollers indicate that the thermal contact between fiber and rollers dominates the heat transfer in the imprinting process. To verify the simulation results, thermal images of PVDF and PEI were captured by thermal camera during the DITD process, as shown in Fig. 2c, d, respectively. The observed temperature distribution in Fig. 2c matches with the simulated results very well. And the temperature distribution of PEI fiber in Fig. 2d follows the similar trend with the simulated temperature distribution shown in Fig. 2b. Also, a temperature drop of 35.5 °C and 51.6 °C is observed after PVDF and PEI fiber is pressed by rollers, respectively.

Based on the simulation and experimental results, we can divide the thermal imprint process into two phases for more detailed discussion. Phase 1 is the surface pattern creating by pressing the template to high-temperature fiber. In this phase, the polymer on fiber surface fills into the cavities on the template to form an inverted pattern under pressure. The polymer flow behavior and important factors affecting the flow behavior are discussed in detail in Supplementary Note 1. As the fiber is drawing down continuously during the DITD process, each point on the fiber surface only contacts with the template for a limited time. Therefore, the filling time must be shorter than the contact time to ensure a high-fidelity pattern. In our case, the contact time is ~50 ms as estimated in Supplementary Note 2, while the filling time can be decided by the following equation31:

$$t = frac{{12eta Sh_{mathrm{c}}^2}}{{Pleft( {2W} right)^2left( {S + W} right)}}$$

(1)

Where (eta) is the viscosity, S and W are the half widths of intender and cavity respectively, hc is the depth of cavity, and P is imprinting pressure. Here, we take the template used in Fig. 1f for example, where S, W, and hc are 2.5 × 10−6 m, 2.5 × 10−6 m, and 1 × 10−7 m, respectively. P is estimated to be 5.7 MPa as discussed in Supplementary Note 2, while a typical viscosity of 106 Pa·s is used. Thus, the filling time can be estimated to be 0.42 ms, which is much shorter than the contact time of 50 ms. Phase 2 is the cooling down of the patterned fiber when contacting with rollers, which helps hinder reflow deformation after the patterned surface is separated from the templates. Figure 2e shows the simulated temperature drop of the fiber being cooled down by the rollers under different drawing temperatures. We can find that the rollers decrease the fiber’s temperature by more than 35 °C in general as the commonly used materials in fiber drawing are always drawn above 150 °C. For semicrystalline polymers such as PVDF and PEEK, such a significant temperature drop will solidify them immediately as the relatively strong intermolecular forces will prevent them from softening below the melting temperature44. Hence, the reflow should be negligible after being cooled down by the rollers. While for most amorphous materials such as PC and PEI, the temperature will directly drop below their glass transition temperature (Tg) after passing through the rollers, thus reflow is also prevented. Only for materials with a relatively low Tg such as PS, the temperature may stay higher than their Tg after being separated from rollers. Then, we can estimate the characteristic reflow time for a periodic surface pattern which is defined as (tau = lambda eta /pi gamma)28, where (lambda) is the period and (gamma) is surface tension. The viscosity will significantly increase when temperature decreases, e.g., the viscosity of PS will increase by 3 orders of magnitude when the temperature drops from 170 °C to 130 °C45. Here we take PS’s viscosity and surface tension as 108 Pa·s and 40 mN/m at 130 °C46. In case of imprinting a pattern with a period of 10−6 m, the reflow time can be calculated to be 800 s. Noticing that fiber’s temperature will keep decreasing under ambient air after being separated with rollers (Fig. 2b), PS’s temperature will drop below Tg in few seconds, which is much shorter than the reflow time of 800 s. Combining phase 1 and phase 2, we can conclude that this imprinting process during fiber drawing is in principle compatible with a wide range of materials and patterns. Additionally, we can always adapt the DITD process to a specific material or pattern by adjusting the parameters including temperature, pressure, and drawing speed.

### Key process parameters Investigation of DITD technique

A resolution test pattern was designed to find the feature size of the DITD technique. The high-resolution template (Supplementary Fig. 8a) was etched by FIB and an inverted pattern was created on the PVDF fiber surface, as shown in Fig. 3a, b. A series of raised rectangular blocks with different sizes and orientations were created. As demonstrated in the enlarged SEM image, the rectangular blocks with a length of 1.5 μm and a width of 500 nm basically maintained their shapes in all directions except for some distortions on edges, confirming a feature size in tens of nanometers. In view of the slight distortion in the template itself (Supplementary Fig. 8b), the actual resolution of the patterns should be better than it appears. Not only the pattern size, the depth of patterns is also controllable. A set of similarly patterned fibers with depths of 2 μm, 8 μm, and 20 μm were fabricated, as the fibers’ cross-section images show in Fig. 3c. Furthermore, various nanopattern arrays were created on fibers to examine the stability and repeatability of the DITD technique, as presented in Fig. 3d–f. Each array included a large amount of rectangular or circular nanoelements with a predesigned period. The rectangle element was 2.1 μm in length and 700 nm in width, while the round one had a radius of 150 nm. All these nanoelements exhibited high fidelity compared with their templates displayed in Supplementary Fig. 9. The high resolution, controllable depth, and high stability of the DITD technique offer surface patterned fibers extensive application areas.

### Plasmonic behavior of nanopatterned fiber

The ability to combine patterned fiber with plasmonics provides opportunities for a variety of applications of fibers, such as biosensing47, bioimaging48, and mode converters49. Here, we investigated the plasmonic behavior of the nanopatterned fibers (with a 30-nm-thick gold film deposited on top) using dark-field spectroscopy. Finite-difference time-domain (FDTD) simulations were performed and the local electric field spatial distributions of the fiber plasmonic systems are shown in Fig. 3g, j which suggest that hot spots were stimulated. The gold pattern models in FDTD were generated from the SEM images as shown in Fig. 3g left and Fig. 3j left. For the nanorod structures indicated in Fig. 3g, the measured dark-field scattering spectra of the plasmonic fiber with transverse (black line) and longitudinal (red line) laser polarizations are presented in Fig. 3h. Two different resonances correspond to the transverse and longitudinal modes of the nanorods. For the nanocylinder arrays indicated in Fig. 3j (pattern 1 and pattern 2), the dark-field scattering spectra with an incidence of non-polarized light are presented in Fig. 3k. Arrays with a smaller period (see Fig. 3j bottom right) possess longer plasmonic resonance which is induced by the near-field coupling. An additional peak near 520 nm is introduced by the edge roughness. The resulting scattering spectra were reproduced by the simulated scattering cross-sections, as presented in Fig. 3i, l. The slight shift of the resonance peaks for both etched patterns when compared with the experimental characterizations can be attributed to the gold layer surface roughness and the minor edge defects, but a good agreement between the shapes of the measured scatterings and the simulated scattering cross-sections was obtained for different patterns.

### Patterned functional fiber-based TENG

Another feature of the DITD process is that it only modifies fiber’s surface structure without influencing its inner structure. This makes it compatible with fibers owning different complex inner structures, predicting a bright future of multifunctional fibers50. TENG is an energy harvesting device that converts mechanical energy including sound, vibration, tide, and wind from surroundings to electricity. It has been widely studied for powering portable devices and self-powered sensing51,52. Previous works53,54 have confirmed that introducing surface pattern can effectively enhance the output performance of TENG. Here, we assembled TENG using surface patterned fiber fabricated by the DITD technique. Figure 4a schematics the fiber’s cross-sectional structure in fiber’s axial direction (left) and displays the optical image of fiber’s top view (right, scale bar, 400 μm). PVDF was selected as cladding material because of its high electronegativity55, while carbon filled polyethylene (CPE) served as the electrode. Both sides of the fiber were patterned with microcylinder array as the SEM images show in Fig. 4b, c. For comparison, the flat-surface fiber with the same dimension was fabricated by the traditional fiber drawing process. Both flat and patterned fibers were cut into 10-cm long pieces. As sketched in Fig. 4d, fiber-based TENG was assembled from 5 pieces of each type of fibers that were connected in parallel (see Methods for more detail). The working mechanism of the single-electrode TENG is discussed in Supplementary Note 3, which indicates that the charge amount on the fiber surface is crucial to the performance of TENG. Creating surface pattern has been proved to be an effective approach to increase the surface charge amount as the surface pattern could increase the surface charge density and enhance the friction53,54. Therefore, the TENG based on surface patterned fibers is expected to offer a better performance than that based on flat fibers. Output performance of fiber-based TENG was measured, as plotted in Fig. 4e, f. The open circuit peak voltage of TENG increases from 2.1 V to 4.8 V due to the presence of surface pattern, while the short-circuit peak current of TENG increases from 160 nA to 500 nA, corresponding to the current density of 0.9 mA m−2 and 2.9 mA m−2, respectively. Moreover, a long-term stability test of TENG based on patterned fibers was conducted as shown in Fig. 4g. The TENG experienced a charge accumulation period at first and then worked stably for more than 40k working cycles. These experimental results indicate that the patterned fiber fabricated by DITD is effective and durable.

### Wearable self-powered multipoint touch sensor

A TENG with higher performance will lead to higher sensitivity and higher signal-to-noise ratio when applied to self-powered sensing applications. Therefore, based on the patterned fibers, we further developed a wearable self-powered multipoint touch sensor. As demonstrated in Fig. 4h, three 10-cm-long patterned fibers and four 8-cm-long patterned fibers were weaved into a piece of cloth and intersected at 12 sensing nodes wrapped around the wrist. When an external object such as fingers, gloves, and metal blocks approaches a certain sensing node, the charge distribution on the electrodes of the two crossed fibers connecting to this node changes because of the electrostatic induction, resulting in voltage output. Hence, the touched point can be electrically illustrated by measuring the output of the 7 fibers. The testing circuit is drawn in Supplementary Fig. 10. Both single-point and multipoint tests were performed by slight finger touches, and the output signals were recorded in Fig. 4i. The numbers labeled on the nodes in the developed view of the wearable sensor (Fig. 4h, right) represent the touching sequence under test. For instance, the first touch occurred on the node intersected by fiber B and E, matching with the result that output voltage was only detected on fiber B and E. Similarly, in the fifth touching test, output voltage was detected on fiber A, B, C, and D simultaneously, indicating a multipoint touching on the leftmost 3 nodes. All the nodes were tested during the 6 touches and the results agreed with touch behaviors very well. Additionally, all the generated voltage from the touch sensor was several volts, which is convenient to be detected.