1. Introduction
With the end of Moore’s Law, integrated circuits have shifted from blindly pursuing an increase in integration to enhancing intelligence [
1]. Intelligent electronics, including smart wearable devices, energy storage devices, sensors, and smart fabrics, are typically adhered to structures or biological surfaces to obtain more accurate sensing signals and enhanced wear comfort, which demands greater flexibility from these circuits [
2,
3,
4]. Additionally, the superb conformal ability of flexible circuits allows them to adhere to the curved surfaces of three-dimensional objects without being limited by installation space. For example, conformal circuits can be applied to radar antennas on aircraft surfaces, structural health monitoring sensors, etc., significantly reducing equipment weight and enhancing aerodynamic performance [
5,
6,
7,
8]. However, traditional silicon-based circuit manufacturing methods, which utilize photolithography, etching, and other processes, are unsuitable for fabrication flexible substrates such as polyethylene terephthalate (PET) and polydimethylsiloxane (PDMS). Screen-printing technology is able to manufacture large-area conductive patterns on flexible substrates and is compatible with high-viscosity pastes, achieving superior conductivity. However, screen printing can only achieve a pattern resolution of at least a few hundred micrometers, thereby posing challenges for the fabrication of high-precision flexible circuits [
9,
10].
The advantages of inkjet printing as an additive manufacturing method are its high precision and environmental friendliness, and it is suitable for various substrates such as metals and flexible polymers [
11,
12,
13]. Among the many inkjet printing methods, electrohydrodynamic (EHD) printing is compatible with various types of nanomaterials, and possesses the ability to prepare flexible electronics with an ultra-fine accuracy below 5 μm [
14,
15]. The principle of EHD printing is to apply a voltage between the nozzle and substrate, and the ink forms a Taylor cone under the action of the electric field that comes into contact with the substrate. At this time, the pathway conducts, the Taylor cone disappears, and then the circuit disconnects to generate a new Taylor cone [
16]. During this cycle, the ink accumulates on the substrate with an ultra-fine resolution, and combined with a precision displacement platform, micro-scale patterns can be printed. Some researchers have prepared flexible heaters, supercapacitors, strain sensors, and other devices through EHD printing [
14,
17]. However, these circuits are almost all single-layer structures, which lead to larger volumes and weights compared to multilayer circuits with the same performance [
18]. Multilayer circuits undoubtedly have more complex preparation processes; researchers use laser etching or photolithography to prepare flexible multilayer circuits, but these methods have high costs [
19,
20,
21]. Therefore, it is necessary to develop a new multilayer circuit manufacturing method to achieve interlayer insulation and interconnection between multilayer circuits.
Bending stress is the most common working load in flexible electronics, which can lead to the cracking, delamination, and degradation of functional materials. Therefore, bending cycle tests are the most important indicator for evaluating the reliability of flexible electronics [
22,
23]. Especially for large multilayer circuits, their higher thickness leads to higher tensile/compressive stress at regions far from the neutral surface during bending, thereby increasing the risk of failure [
24]. For circuits prepared via inkjet printing, it is difficult to ensure sufficient adhesion between the conductive layer and the insulation layer, and delamination may occur during the bending cycle process, which poses challenges to the selection of conductive ink and insulation layer materials. Given this research background, this study adopts a new flexible circuit fabrication method based on EHD printing, aiming to prepare large-area flexible multilayer electronics with excellent bending reliability. In order to make multilayer circuits more compatible with a wider range of usage scenarios, thermal cycling reliability is also considered.
3. Results and Discussion
In high-precision printing, the line width is an important indicator for evaluating the printing effect. The width of the conductive path is also an important design parameter, as uncontrollable line widths would lead to short circuits between pins. Unlike traditional lithography processes, printing electronics cannot obtain specific line widths through masks, and the line widths acquired through printing have a strong process correlation. Therefore, it is necessary to clarify the rule of variation in line width through various process parameters to obtain conductive patterns with specific line widths. The influence of the printing speed, working height, and nozzle diameter of EHD printing on the line width is shown in
Figure 3. As the printing speed increases, the line width gradually decreases, and the rate of change of the line width gradually decreases before stabilizing. The printing speed does not affect the diameter of the Taylor cone jet, but it does determine the amount of ink deposited per unit of time. A lower speed leads to greater ink deposition, and the wetting and spreading of the ink on the substrate increase the line width, as shown in
Figure 3a. As the working height increases, the line width steadily increases due to the diffusion of the Taylor cone jet in the air, as shown in
Figure 3b. When the working height approaches zero using high-speed printing, the line width becomes close to the diameter of the Taylor cone jet. The diameter of the nozzle directly determines the diameter of the Taylor cone jet. As the nozzle size increases, the line width steadily increases, as shown in
Figure 3c. When the nozzle diameter is too large, it cannot produce a stable Taylor cone jet. Moreover, excessive printing speeds and working heights lead to discontinuous printing lines, resulting in large edge roughness of the printed lines. The diameter of the drill bit used for drilling in this study is 150 μm. The drill bit damages the conductive path when the line width is too small. We purposefully selected a parameter combination based on the designed line width of 200 μm, so we used a 300 μm nozzle, a working height of 100 μm, and a printing speed of 0.5 m/s in the subsequent multilayer circuit preparation process; using these parameters, we were able to reliably obtain a 200 μm line path, as shown in
Figure S4 of the Supplementary Materials. The deposition rate of the ink is also a key parameter, and we calculated the ink deposition rate of 1.2 mL/h.
Generally, metal nanoinks require post-treatment above 150 °C to exhibit good conductivity. However, the glass transition temperature of PET substrates is approximately 110 °C, so post-treatment needs to be carried out below 110 °C. In addition, we need to determine the conductivity mechanism, which is necessary to explain the increase in resistance during the bending cycles. The morphologies of the printed conductive path after 60 min of sintering at different temperatures are shown in
Figure 4a–d; combined with the electrical resistivity results of
Figure 4e, it can be seen that after post-treatment at temperatures below 150 °C, there was no significant change in the microstructure of the silver path, and micrometer sized silver nanoflakes were uniformly mixed as shown in
Figure 4a–c. After sintering at 100 °C for 1 h, the resistance of the conductive circuit showed a significant decrease compared to the one without post-treatment, with a resistivity of 14.3 μΩ·cm. When the post-treatment temperature was raised to 150 °C, the resistance did not decrease further, indicating that the decrease in resistance caused the volatilization of organic solvents, which led to a decrease in the contact resistance between nanoflakes. However, at this time, these sliver nanoflakes were not sintered. When the temperature was raised to 300 °C, the resistance significantly decreased and quickly stabilized. The nanoflakes were sintered together due to the atomic diffusion at such a high temperature, which further reduced the resistivity, as shown in
Figure 4d. In summary, sufficient conductivity could already be obtained at 100 °C via organic volatilization and the overlap between silver flakes, so we put the printed ten-layer circuit in a constant temperature box at 100 °C for 1 h.
Based on the process above, the ten-layer flexible circuits were prepared as shown in
Figure 5. After testing the conductivity between different pins, effective interconnection was achieved between each layer. The unconnected pins were non-conductive, which proved that there was no short circuiting. The purpose of using a flexible substrate is to attach the circuit to a three-dimensional surface, so we attached the ten-layer circuits onto a wing-shaped resin substrate, as shown in
Figure 5b,c. The resin block was fabricated using photocurable 3D printing, and the minimum curvature radius of the curved substrate was 20 mm. The flexible multilayered circuit showed an excellent conformal ability and tightly adhered to the curved substrate without delamination, proving that the multilayered circuit can adapt to complex curved surfaces.
Bending stress is the most common load on flexible circuits. To improve the bending reliability of the multilayered circuits, we established different conductive layer thicknesses in different areas in order to improve it. The part above the neutral plane bears the tensile stress during the bending process, while the part below the neutral plane bears the compressive stress. As mentioned in
Figure 4, the good conductivity at a post-processing temperature of 100 °C is due to the overlap between the nanoflakes rather than a result of sintering. Compression stress will compress the nanoflakes to enhance conductivity. Therefore, the conductive layer can be printed thinner in the areas with higher compression stress. Conversely, tensile stress will loosen the nanoflakes, and an excessively thin conductive layer may result in an open circuit. Therefore, we made the conductive layer thicker in areas with higher tensile stress. We achieved a thicker conductive layer by increasing the number of prints, as shown in
Figure 6a–c. We repeated printing twice for the area of the 10-layer circuit under compressive stress and obtained an average thickness of 5.6 μm. For the area under tensile stress, we repeated printing five times and obtained an average thickness of 26.5 μm, as shown in
Figure 6d. A cross-section SEM image of a conductive through-hole is shown in
Figure 6e. It can be seen that the conductive through-hole connected different conductive layers together. The interface between the conductive and insulating layers was tightly bonded without any delamination phenomenon, as shown in
Figure 6f. EDS surface scanning was performed on the ten-layer circuit’s cross-section, and the results are shown in
Figure 6g–i. The main component of the conductive layer was Ag, and O and C elements were concentrated in the polyimide insulating layer, which separates the different conductive layers well and avoids the formation of short circuits. In addition, it can be found that there was no Ag in the insulating layer, but there was a low concentration of C and O elements in the conductive layer, which is due to the fact that we used a post-processing temperature of 100 °C and the organic compounds in the ink did not evaporate completely.
The curved substrates result in the flexible circuit undergoing bending stress for a long time. In order to verify the reliability of the flexible multilayer circuit under cyclic bending with a large curvature radius, a combination of simulations and experiments was carried out to explore the reliability and failure mechanism of the multilayer circuits. The displacement distribution and stress distribution are shown in
Figure 7a,b, indicating that stress is concentrated in a small range of the bending center, only 0.55 MPa, due to the very small modulus of PET. A bend cycling test was conducted on flexible circuits to monitor the changes in resistance during the bending process, as shown in
Figure 7c. The resistance exhibited periodic changes with increasing amplitude over the bending cycle. After 500 cycles, the resistance change rate was 7.9%, and after 8000 cycles, the resistance was approximately three times the initial resistance, demonstrating excellent bending reliability. Results of the microscopic morphology analysis of the circuit after 8000 bending cycles are shown in
Figure 8. It can be seen that the microstructure on the surface of the circuit did not show obvious changes, but fine cracks appeared on the surface with a width of 1 μm. During the cyclic bending process, the maximum crack width was observed when the bending angle reached its maximum value and led to the maximum resistance. As the bending angle decreases, the nanoflakes around the cracks overlapped with each other, reducing the resistance and exhibiting periodic changes. Additionally, we found that, after the bend cycling experiment, the resistance of the circuit continued to slowly decrease. After 8000 bending cycles, the resistance was 2.6 times the initial resistance. However, the resistance decreased to 1.7 times the initial resistance after 10 h. This may be due to the effect of gravity, which makes the loose nanoflakes overlap in a tighter configuration.
In order to make the multilayer circuit more compatible with a wider range of usage scenarios, the temperature cycling test environment was taken from −25 °C to 85 °C, and corresponding thermal cycling simulation was carried out. Conformal electronics experience both bending stress and thermal stress. Therefore, thermal–mechanical coupling simulations were carried out with bending prestress, and we applied the stress distribution obtained from bending simulation as a load, which can be used to explore the thermal cycling reliability of multilayer circuits under bending conditions. The temperature distribution at the moment of maximum temperature gradient during the heating process is shown in
Figure 9a. The temperature distribution of the entire multilayer circuit was consistent, with a maximum temperature difference of only 0.13 °C. This is because the thickness of the multilayer circuit was only 0.2 mm, and the temperature quickly reached uniformity in a hot convection environment. The maximum stress during the thermal cycling process occurs at extremely low temperatures because the difference in the thermal expansion coefficient of the material is the greatest. Thermal stress is generated by mismatches in temperature, and the maximum stress point occurs at the junction of the silver line at the bending center and the PET, which was 6.8 MPa. It can be seen that the thermal stress generated by this multilayered circuit under thermal cycling load is extremely small. Due to the long testing time and the large amount of data required to detect changes in circuit resistance, multiple samples were tested simultaneously, and the initial resistance was measured as R
0. Samples were measure under 100, 200, 500, 800, and 1000 temperature cycles for resistance measurement, and the measured value was denoted as R’. The resistance change during thermal cycling is shown in
Figure 9c. After 200 cycles, the resistance dropped to 95% of the initial resistance and remained unchanged. The initial slight decrease in resistance was because of the annealing effect in the high-temperature region, which further densified the conductive path. The experimental results were consistent with the simulation results, and the minimal thermal stress caused by thermal cycling did not cause a decrease in performance. Moreover, the resistance remained stable after 1000 thermal cycles.