3.1. Results of Metal Transfer Condition
The formation of a strong bonding between PDMS and copper can be observed as a transferred copper layer on PDMS. The copper layer was transferred from the chromium layer, which acts as an adhesive layer between the glass substrate and a copper layer for easier handling during photolithography and wet etching. Without an adhesive layer, the target metal can be easily damaged. Moreover, it is evidence to show that our method creates strong bonding between the metal and PDMS.
The theory of our method in transferring a metal layer onto PDMS is creating stronger adhesion between the target metal (copper) and PDMS than adhesion between the target metal and adhesive layer. If the adhesion between PDMS and the copper layer is stronger than the adhesion between the adhesive layer and the copper layer, the copper layer can be transferred onto PDMS. The standard PDMS curing condition cannot generate strong bonding between any metal and PDMS, and, as a result, the standard condition cannot transfer the copper layer onto PDMS. The standard condition of curing PDMS is baking a PDMS of 10:1 mixing ratio at 65–95 °C for 15–120 min. We modified the PDMS curing conditions to create a strong bonding between PDMS and the copper layer by PDMS curing process as shown in
Figure 4 and
Table 1.
In the case of using the PDMS of 20:1 and 10:1 mixing ratio, no copper transfer was observed even though the baking temperature was high and the baking time was long. For PDMS of a 5:1 mixing ratio, the transferred copper layer was observed. The results showed that it was possible to transfer copper with a low baking temperature (100 °C). However, in the case of the low baking temperature, the baking time must be longer than 20 min for initiating the copper transfer and only a small amount of the transferred copper was observed (partial transfer). The percentage area of transferred copper was analyzed using ImageJ software (NIH, Bethesda, MA, USA) by comparing the area of transferred copper and the total area of the copper pattern. When the PDMS was baked for 30 min, ~4% of the copper layer was transferred. When the PDMS was baked for 45 min, about 6% of the copper layer was transferred. This shows that the critical conditions for generating a strong adhesion between copper and PDMS are a greater amount of curing agent mixed in the PDMS mixture (5:1 mixing ratio), high baking temperature (150 °C) and longer baking time. Additionally, when the PDMS of 5:1 mixing ratio was baked at 150 °C for more than 20 min, the PDMS started to bond with chromium and PDMS was broken while peeling off. The chromium layer is initially deposited on a glass substrate prior to copper deposition. After copper etching, parts of the chromium layer are exposed to air (or PDMS in the PDMS curing process). This indicates that it is possible to create bonding selectively by controlling baking time.
This was a simple comparison between standard PDMS curing conditions and our modified curing condition in the creation of strong bonding. As the result shows, the condition required for generating strong bonding between copper and PDMS is a baking PDMS mixing ratio of 5:1 at 150 °C for 20 min. In addition, we expect that our method is compatible with various metals by modifying the PDMS curing conditions such as increasing baking time.
The science of creating strong bonding by changing PDMS curing conditions is not yet fully understood. It requires an extensive amount of study, especially related to surface modification caused by our PDMS curing condition.
3.2. Bonding Strength Test Results
The cross-hatch adhesion test was repeated five times and the results are shown in
Figure 5. According to the ASTM adhesion strength class, the cross-hatch test is divided into six classes, 0B–5B. 5B indicates that the removed area is 0% and 0B indicates that the removed area is greater than 65%. Due to the flexible nature of the PDMS, the distance between some cuts was not uniform. Therefore, we analyzed the area in the red box where the cuts are uniform. As shown in the
Figure 5a, no copper was removed by the tape (0.1% ± 0.03%, area analyzed by ImageJ software). This indicates that the ASTM class of copper transferred by our modified method is 5B. In the case of the transferred copper by standard condition, more than 65% area was removed by the tape (82% ± 4%, area analyzed by ImageJ software), which means that the ASTM class is 0B. This proves that our method had successfully improved the adhesion strength between a copper layer and PDMS.
If the copper layer is deposited on a glass substrate without an adhesive layer, it is possible to transfer copper onto PDMS by the standard curing condition without any chemical treatment. However, this will result in a weak adhesion between copper and glass substrate, which, in turn, can be peeled off with great ease [
17]. This is why the adhesion strength of transferred copper using standard condition is classified as 0B.
3.3. Result of Optimum Electrode Thickness
The sheet resistance of various copper thicknesses was measured 10 times sequentially and the values were averaged. As shown in
Table 2, the sheet resistance was increased after the transfer, if the copper thickness was ≤150 nm, but, if the copper thickness was ≥300 nm, the sheet resistance decreased after the transfer. Based on the results, the rate of sheet resistance decreased at a copper thickness of 500 nm. Based on the experimentation, it can assumed that the optimum copper thickness is approximately 500 nm after peeling.
Initially, we expected that the sheet resistance of the electrode (of any thickness) would increase after transfer due to the cracks and wrinkles on copper surface during the PDMS peeling process. The increase rate of sheet resistance would be inversely proportional to the thickness of the electrode. However, the result showed that the sheet resistance decreases after the transfer if the thickness of copper is ≥300 nm.
To study the relationship between metal thickness and the amount of cracks and wrinkles generated on the surface, a series of SEM images of each copper thickness were taken as shown in
Figure 6. The images show that more wrinkles and cracks were present on the surface of the thinner copper. The surface of the 75 nm thick copper electrode was the roughest and the surface of the 1000 nm thick copper electrode was the smoothest. If the surface smoothness caused by the creation of cracks and wrinkles is the major cause of change in sheet resistance after transfer, then the sheet resistance of 1000 nm must have the greatest decrease after transfer. However, the result in
Table 2 shows that the 500 nm thick copper electrode has the greatest decrease rate of the sheet resistance after transfer. This result is similar to the study of Lu et al. in 2009 [
15], which showed the effect of the copper thickness on the failure strain and that the 500 nm thick copper electrode was more resistant to the strain. In their study, they calculated the yield strength of different copper thicknesses using the Hall–Petch relation
σy =
σ0 +
kd−1/2 [
15,
18,
19] where
σy is the yield strength,
σ0 is the material strain constant coefficient of starting dislocation motion, k is the constant strengthening coefficient and
d is the grain diameter. The Hall–Petch relation indicates that the yield strength of metal film is proportional to
d−1/2; therefore larger grain size will lead to lower yield strength of the copper electrode. The grain size is highly dependent on various conditions such as metal deposition rate and temperature, but, according to the study of Dammers and Radelaar [
20,
21,
22,
23,
24], and the grain diameter is directly proportional to square root of film thickness. Since the copper deposition condition for every thickness was identical, we can assume that the thicker copper layer would have larger grain size. In the study of Lu et al. in 2009 [
15], the deposition condition for all copper thickness was identical and the grain size was increased as the thickness of copper electrode was increased. Similarly, our copper deposition condition was identical for all thickness, therefore we assume that the thicker copper electrode would have a smaller yield strength, which results in less durability of, in this case, the 1000 nm thick copper electrodes.
According to the Hall–Petch relation, a thin copper electrode (<200 nm) has a greater yield strength, which means that it can endure more stress. However, it experiences intergranular fractures as stress is applied; hence, more cracks are generated (rougher surface) and the sheet resistance is increased significantly. On the other hand, an electrode thickness above 200 nm experiences transgranular fractures from yield strength and less cracking is generated. This leads to the smoother surface of copper electrode allowing it to endure greater stresses [
15].
A topic of further exploration for this research is the investigation, in terms of analytical data and explanation, of the decrease in sheet resistance of the transferred copper thickness is above 300 nm. Based off the current research, the strongest explanation for this is due to the decrease of the electrode length during the thermal expansion and compression. Applying a high temperature (150 °C) is the key of creating a strong bonding without any chemical treatment in our method, which causes thermal expansion PDMS when heated and then compression when cooled. According to the study of Bowden et al. [
25,
26], the PDMS is expanded by heat (thermal expansion), which creates wrinkles as it cooled down (compressed). The cracks on the copper surface suggests that there could be a thermal expansion mismatch between the copper and PDMS (additional shrinking in polymerization). The thermal expansion coefficient of copper is 17 × 10
−6 °C
−1 and PDMS is 3 × 10
−4 °C
−1 [
25,
26,
27,
28,
29]. The large difference in thermal coefficient between two materials cracks and wrinkles the surface of copper after cooling down. The surface of PDMS bonded to metal would not be expanded as much as the other side; however, the thermal expansion of the surface of the PDMS bonded to the metal would be enough for the creation of wrinkles when it compressed. As a result, the length of the copper electrode was decreased, which led to the decrease of sheet resistance after transfer (
R =
ρl/A, where
R is resistance,
ρ is resistivity,
l is the length of the electrode and
A is the cross sectional area of the electrode). The difference in the decrease rate is caused by the fracture mechanism of the 500 nm thick copper being more resistant to stress than that of the 1000 nm thick copper [
15]. Thus, the decrease rate of the 500 nm thick copper is greater than that of the 1000 nm thick copper.
3.4. Result of the Resistance Change Flexible Electrode
The change in resistance by bending the rectangular flexible copper electrodes using the cylinders of various radii was measured and shown in
Figure 7a. The resistance change was measured 10 times repetitively and then the results were averaged. The results show that the 500 nm has the least increase of the resistance. The original resistance of 300 nm, 500 nm and 1000 nm thick copper electrode were 0.88 Ω, 0.4 Ω and 0.28 Ω, respectively. The resistance change of the 300 nm, 500 nm and 1000 nm thick copper electrode after the bending using the cylinders of various radii were 3.02 (
R/
R0), 2.24 and 2.39, respectively. The result shows that the resistance increase of the 500 nm thick copper electrode was the most stable and the 300 nm thick electrode shows the greatest resistance increase. As mentioned in the study of Lu et al., the 500 nm thick electrode has the least increased rate of resistance, which is lower than the 1000 nm thick electrode because strain to rupture 500 nm is better than that of the 1000 nm due to the fracture mechanism.
The result of the change in resistance of the flexible electrode by repetitive bending using a cylinder of radius of 11 mm is shown in
Figure 7b. The result shows that, as the thickness of the copper electrode increases, the rate of the resistance increase is smaller. The resistance of the 300 nm thick copper electrode after 1000 bending cycles was increased to 8.37 (
R/
R0) times greater than its original value (0.91 Ω), and it seems that it would increase after more bending cycles. The resistance of the 500 nm thick copper plate at 1000 bending cycles was increased to 3.37 (
R/
R0) times greater than its original value (0.45 Ω), though it seems that the resistance stayed constant after 900 bending cycles. The resistance of the 1000 nm thick copper plate at 900 and 1000 bending cycles was increased to 2.27 (
R/
R0) times greater than its original value (0.32 Ω). 600 bending cycles and fewer the resistance stayed constant.
The R/R0 value of the 500 nm thick copper electrode was lower than the 1000 nm thick copper electrode below 400 bending cycles, which follows a similar trend as the bending radii test. However, the R/R0 of the 500 nm thick copper electrode increased rapidly beyond 400 bending cycles, and it became higher than that of the 1000 nm thick copper electrode. This result shows that the thicker electrode is more stable against repetitive bending. The maximum stress that the electrode can endure depends on the thickness of the electrode and its grain size (which is proportional to the thickness of film). Thus, in selecting the thickness of the electrode, the optimum thickness must be selected for stability against the deformation of flexible electrode as well as choosing the right material for the electrode.
Despite the advantage of our method being simple and cheap, the defect and deformation of electrode created during the transfer process limits our method. Cracks are generated as shown in the SEM images in
Figure 6 due to the mismatch of thermal expansion coefficient [
25,
26,
27,
28,
29]. The crack and wrinkles cannot be avoided at this point; however, there are a few strategies to overcome this limitation. One strategy is applying the serpentine electrode design [
30,
31,
32,
33]. The serpentine design mechanically enhances the metal resistance to cracking; hence, we will design electrodes before transfer through photolithography for better quality in future studies.
3.5. The Result of Pressure Sensors
The strain can be calculated by using the equation,
C =
εoεA/
d (where
εo is dielectric constant of free space,
ε relative dielectric constant of the material, A is area of electrode,
d is the distance between each electrode). The result of measured strain is shown in
Figure 8. The graph was obtained by averaging values from five repetitive experiments, and the slope was obtained by drawing a trend line using Microsoft Excel (Microsoft Excel 2013, Redmond, WA, USA). The 1/slope of the graph is Young’s modulus (Young’s modulus = pressure/strain) of dielectric material of the pressure sensor. The dielectric layer of the pressure sensor is composed of three different materials: air, PDMS (15:1) and PDMS (5:1). These materials have a different Young’s modulus, PDMS (15:1) = 1.4 MPa and PDMS (5:1) = 3.59 MPa [
34] (air does not have Young’s modulus because it is not solid). Due to the difference in the Young’s modulus of each material, we expect that each material would deform at different applied pressures (three ranges of pressure because there are three materials composing the dielectric layer). The strain change can be divided into three ranges: low pressure range (1–289 Pa), middle pressure range (289 Pa–10.408 kPa), and high pressure range (10.408–131.81 kPa).
Unlike our expectation in designing the pressure sensor, each range seems to correspond to the deformation of the combination of materials. In the low pressure range, the air gap was deformed and its Young’s modulus from the graph is 5000 Pa (the slope is ~2 × 10−4). This does not mean that air has Young`s modulus, but, as the air gap was fabricated within the PDMS (15:1), the calculated value would require pressure to deform the air gap with the PDMS (15:1) (the main deformation would be generated by air). In the middle pressure range, the PDMS (15:1) that has lower Young’s modulus was deformed and the calculated Young’s modulus from the graph is ~1 MPa (the slope is ~1 × 10−6). The theoretical Young’s modulus of PDMS (15:1) is 1.4 MPa, which is greater than our value. This would be caused by the existence of an air-gap that decreased the Young’s modulus of the middle pressure range (the PDMS (15:1) was mainly deformed). In the high pressure range, PDMS (5:1) was deformed and the calculated Young’s modulus from the graph is ~2.5 MPa (the slope is ~4 × 10−7). The theoretical Young’s modulus of PDMS (5:1) is 3.59 MPa, which is greater than our value because the deformation by the high pressure range was done by a combination of the PDMS (15:1) and the PDMS (5:1).
Due to the combined material deformation, further studies for optimization would be required to test in a practical experiment. Moreover, further study is required for the demonstration of a flexible pressure sensor because, when the pressure sensor is bent, the capacitance change is unstable. At this time, we demonstrate a possible application of our method through fabricating flexible pressure sensors.
3.6. The Result of Different Metal Transfers
The chemical-free metal PDMS thermal bonding is compatible with nickel and silver as shown in
Figure 9a. As the result shows, the transfer of a micro-size pattern of nickel and silver from the adhesive layer (chromium) was successful using our method (as well as copper). Additionally, our method can transfer the micro-size pattern on PDMS with the widths of silver and nickel 400 µm, 300 µm, 200 µm and 100 µm easily transferred on PDMS. The possibility of transferring micro-size patterns of nickel allows fabrication of flexible soft magnets and soft magnetic structure integrated microfluidic devices. The soft magnetic structures in microfluidic devices can be used in magnetic field based particle sorting or bioassay [
35,
36]. Additionally, the possibility of transferring micro-size patterns of silver allows fabrication of sensors as well.
The result of the cross-hatch adhesion test of nickel and silver is shown in
Figure 9b. The ASTM class of nickel adhesion to PDMS is 5B because 0% nickel was removed by the tape. The ASTM class of silver adhesion to PDMS is 0B because more than 65% silver was removed by the tape. The result shows that the adhesion strength between nickel and PDMS is strong; however, the adhesion between silver and PDMS is weak. It seems that the adhesion strength between silver and PDMS is not stable (mostly weak but strong adhesion can be generated at low success rate) and transferred silver is easily damaged. This indicates that the silver transfer requires more optimization, which requires further studies.
Similar to cracks on transferred copper, there are cracks on transferred nickel and silver. This is also caused by the thermal expansion mismatch and the same strategy for copper (such as serpentine design) can be studied in the future to avoid cracking.
There are other metals that are compatible with our method such as aluminum, chromium and permalloy. However, the transfer of these metals is difficult because the condition is not optimized. This process is still developing; therefore we believe that further studies of this method can enable the transfer of aluminum, permalloy and other metals.