Highly Stretchable Transistors Using a Microcracked Organic Semiconductor

Organic transistors with elastic conductors and dielectrics can be stretched up to 250% strain while maintaining the transistor characteristics. Strain-independent properties can be achieved after an initial “programming” cycle that causes the formation of microcracks in the semiconductor. The change in mobility with strain follows the same trend in different stretching directions. Liberating electronic devices from the confines of traditional rigid substrates can improve mechanical robustness and enable new applications and manufacturing methods. Stretchability facilitates electronics that can be mounted on unconventional substrates,1 such as lenses and human bodies,2 and allows dynamic tuning of devices such as electronic eye cameras3 and lasers.4 Accommodating complex movements of supporting structures facilitates integration with moving entities and is critical for biointerfacing applications5 and electronic skins6-10 for prosthetics and robotics. Arrays of electronic devices often include transistors as active addressing elements in order to improve the signal collection process.5, 8 Furthermore, many applications, including sensor arrays6, 11 and displays,12 require large area coverage and therefore benefit from low-cost, high-throughput fabrication methods. In this communication, we report a stretchable organic transistor that maintains transistor behavior to >250% strain, which is several times larger than previous reports.13-16 Strain-independent characteristics are achieved by “programming” the device with an initial strain that causes the formation of microcracks in the semiconductor layer. Crack formation accommodates strain, while maintaining a percolating pathway. Similar microcracking17, 18 or void formation19 strategies have been employed successfully in stretchable conductors used in applications such as strain sensors19 and neuroprosthetic devices.20 The fabrication process involves cost efficient solution methods including spraycoating and spincoating. Stretchable electronics can be fabricated using two main methods. The first involves geometrical patterning of conventional electronic materials such as metals and inorganic semiconductors into meandering patterns or buckles in order to reduce deformation in the active material.21-24 Stiff islands connected with stretchable conductors have produced arrays of high-performance devices including transistors,25 photodetectors,3 and LEDs.12 However, because stretchability is imparted by the regions between islands, there is a trade-off between device density and stretchability. Buckling involves the formation of wavy structures that flatten out to accommodate applied strain. The optical properties of these wavy structures could benefit the performance of stretchable solar cells,2, 24 but may be undesirable for other optoelectronic applications,26 or for devices that require planar interfaces. The second method of imparting stretchability is to fabricate devices composed of elastic materials. Because elastomers are typically insulators, electronic functionality is often imparted by blending with electronic materials12, 26 or applying thin films of compliant electronic materials.15 Several intrinsically stretchable electronic devices have been reported, including stretchable light emitters based on electrochemical active layers26, 27 and graphene-13 and MoS2-based14 transistors that stretch to 5%. A hybrid method was reported by Chae et al. in which an intrinsically stretchable graphene-based gate electrode was combined with a buckled inorganic dielectric layer to make high-performance transistors that could sustain repeated strain cycles to 20%.15 Inorganic semiconductors with proper fabrication schemes and device design can provide exceptional performance in stretchable electronic devices,3, 28 but their high processing costs limit implementation in applications where devices need to be disposable, cheap, or cover large areas. Organic semiconductors (OS) are an alternative that have been touted for their materials availability and compatibility with high-throughput, room-temperature deposition methods such as slot die coating29 and inkjet printing.30 While the performance of OS has traditionally lagged behind that of their inorganic counterparts, their electrical characteristics have been steadily improving,31 and their mechanical properties may be more suitable for compliant electronics. The strain tolerance of OS can vary; research by O'Connor et al. suggests that a rigid, 3D packing structure results in a very low strain at fracture, while the 2D packing structure of a common polymeric semiconductor poly(3-hexylthiophene) (P3HT) allows deformation to greater than 150%.32 Consequently, P3HT was chosen as the semiconductor in our stretchable transistors due to favorable mechanical characteristics, well-characterized properties,33 and ready availability. Devices were fabricated in a bottom contact, top gate architecture, as depicted in the schematic in Figure 1a. The source and drain (S/D) electrodes were composed of carbon nanotubes embedded in a polyurethane elastomer (PU), similar to electrodes first published by Pei and coworkers.27 The sheet resistance of the electrodes before and after embedding in PU was ∼155 Ω/sq and 945 Ω/sq, respectively. P3HT films were transferred from a wafer coated with a hydrophobic silane monolayer34 onto the embedded CNT electrodes, and a 4 μm thick PU dielectric layer (Figure S1) was spincoated on top. The device was completed by applying eutectic gallium indium (EGaIn), a liquid metal, as the gate electrode. Figure S2 provides microscope images of a device at different stages throughout the fabrication process, and a more thorough description is given in the experimental section. Figure 1b and 1c depict the device at 0% and 150% strain. The electrical characteristics of the assembled transistors were collected using a probe station in a nitrogen environment, and a typical transfer curve is provided in Figure 1d. The average and standard deviation of the transistor characteristics include mobility (μ) values of μ = 3.4•10−2 ± 1.63•10−2 cm2/Vs and an on/off ratio of 591 ± 461. The large variation in device characteristics was a consequence of the manual fabrication processes, which included transferring the semiconductor, spincoating the dielectric, and applying the top gate. The device performance is expected to be closely related to the nature of the dielectric. A large dielectric thickness and concomitant low capacitance (∼1 nF/cm2) can lead to impaired mobility values, due to a lower applied electrical field. Correspondingly, the on/off ratio was also lower than some other reports.35 Transfer characteristics of a device stretched to 160% and released back to 20% are provided in Figure 1d. Additionally, exposure to air is known to result in a lower on/off ratio for P3HT transistors due to oxygen and water doping.36 Furthermore, the loose packing and flexibility of PU results in a tendency to absorb polar solvents, which can have several potential effects on the device performance: (1) solvents can disrupt charge transport, reducing charge mobility13 (2) absorbed water increases the polarization in the dielectric and consequently the measured on current (ION), distorting mobility values,37 and (3) polar groups can increase leakage, off currents (IOFF), and hysteresis.38 While there is much room for improvement in dielectric selection, the main purpose of this report is to analyze the relative performance of the device with applied strain. The changes in P3HT morphology with strain were investigated using three structures: (1) P3HT transferred onto a PU substrate (PU/P3HT-t) (Figure 2a-c), (2) PU Spincoated onto P3HT and then transferred onto a PU substrate (PU/P3HT-s) (Figure 2d-f) and (3) (PU/P3HT-t) with a dielectric spincoated on top (PU/P3HT/PU) (Figure 2g-i). In the PU/P3HT-t structure, optical microscopy revealed that cracking started in the semiconductor at strains less than 15% strain. At 65% strain, large cracks with widths in the range of 10 μm were observed (Figure 2b). In addition to the microscale cracks observable with optical microscopy, smaller nanoscale crack-like defects could be resolved using AFM (Figure S3). Continued stretching increased the crack width (Figure 2c). With PU/P3HT-s, the onset of observable microscale cracking occurred between 40% and 65% (Figure 2e). Similar to the PU/P3HT-t structure, the crack length and width increased with strain (Figure 2f). However, the widths of the cracks in the PU/P3HT-s structure were in the range of hundreds of nanometers to several microns. The improved stretchability may be due to a better adhesion of P3HT on the spin coated PU inhibiting crack formation in the semiconductor. Observations for PU/P3HT/PU were similar to those from PU/P3HT-s. The crack size in this structure is much smaller than the device dimensions, indicating that crack formation does not induce substantial variability between devices. The orientation of polymer backbones can be investigated using UV-Vis measurements polarized perpendicular and parallel to the stretching directions. The dichroic ratio (R) is defined as the ratio of the peak intensity of the absorption polarized parallel to the stretching direction divided by the peak intensity of the perpendicular absorption (R = A∥/A⊥). When strain is accommodated by continuous plastic deformation in P3HT, alignment of the polymer chains in the direction of stretching causes anisotropy in the optical absorption, resulting in a dichroic ratio larger than 1.39 The trend in R with strain displayed key differences between the different structures with P3HT (Figure 2j). The PU/P3HT-t sample exhibited a slight increase in R at small strains followed by a gradual return to 1. This suggests that a small amount of plastic deformation is accommodated before cracking occurs. Subsequent stretching resulted in misorientation of the partially-aligned domains, causing R to approach 1. In contrast, the PU/P3HT-s and PU/P3HT/PU structures exhibited a linear increase in R up to ∼1.3 at 50% strain, followed by a region with relatively little change in R. The UV-Vis observations are consistent with those from optical microscopy; both characterization methods indicate that large-scale cracking begins below 15% strain for P3HT transferred onto PU (PU/P3HT-t) and in the range of 50% strain for the systems in which spincoating was used (PU/P3HT-s and PU/P3HT/PU). Adhesion of a plastic material to a deformable substrate improves ductility by limiting the localization of strain that is responsible for crack formation.40 The spincoating process may facilitate excellent adhesion by conformally coating the P3HT, limiting delamination. UV-Vis spectra collected before and after spincoating the dielectric (Figure S4) suggest that the semiconductor film is largely unchanged. Several reports have described the continuous deformation of P3HT to large strain values without the formation of cracks. However, these reports used high temperatures during stretching41 or higher molecular weight P3HT with improved adhesion promoted by UV/Ozone treatments.39 UV/Ozone treatments were detrimental to the device performance in this work because of doping effects that increased IOFF. Electrical measurements were collected while stretching the devices both perpendicular and parallel to the direction of current flow in the channel (Figure 3a). When stretching was undertaken parallel to the charge transport direction (perpendicular to the orientation of the S-D electrodes), the ION decreased rapidly. The experiment was concluded at 140% strain when the On/Off ratio reached below 10. Compared to the parallel direction, the ION decreased more slowly while stretching in the perpendicular direction, and the devices exhibited transistor characteristics to ∼265% strain (Figure 3a). These uniaxial strain values are significantly larger than reported in previous work.13-16 The output characteristics displayed in Figure S6 showed that there is little contact resistance in the measured strain range. Contact resistance is discussed more thoroughly in the supplementary. The gauge factor (GF) is a parameter that is commonly used to quantify the sensitivity of strain sensors. GF is defined as (ΔR/R0)/ε, where ΔR is the change in resistance, R0 is the initial resistance, and ε is the strain. A small GF is sought for application in stretchable active matrices. In the parallel stretching direction, the GF began at ∼7 at low strains, and slightly decreased before increasing at high strains (Figure 3b). In contrast, devices stretched in the perpendicular direction exhibited a GF close to 2 throughout the measured strain range. These values compare favorably to a GF of >10 estimated from published data on stretchable graphene transistors.13 The mobility of the devices decrease at a similar rate for both the perpendicular and parallel stretching directions (Figure 3c). The method of calculating mobility is included in the supplementary information. The increase in leakage current (Figure 3d) is consistent with a reduction in the dielectric thickness. The threshold voltage does not show a stable trend with strain (Figure S7). A strain series was conducted in the parallel direction (Figure 3e), which shows reversible characteristics at small strains, with the degree of irreversible changes increasing with strain. Strain-independent ION was achieved during multiple perpendicular stretching cycles (Figure 3f); ION decreased by ∼50% during the first cycle and remained relatively constant throughout the subsequent cycles. Similar strain-independent properties after an initial prestretch have been observed in CNT-based stretchable electrodes6 and in pentacene transistors.42 Reversible changes in OS transistors have been observed only at low strains,43, 44 which is consistent with our observations. In both stretching directions, IOFF was limited by leakage through the dielectric, which changed modestly with strain (Figure 3c). Consequently, the trend in the on/off ratio followed ION. The collected data facilitate the consideration of possible mechanisms for strain-dependence in the electrical properties. Microscale cracking was not observed until ∼50% strain, indicating that the rapid reduction in ION at low strain values must be attributed to other processes. One possible process could be the formation of nanoscale cracks observed by AFM in Figure S3, which could not be resolved using optical micro­scopy. Alternatively, work on flexible devices has suggested that strain sensitivity can arise from both changes in the separation between molecules44, 45 and changes in the separation between crystallites.43, 46 The relatively constant shape of the UV-Vis spectra with strain (Figure S9) suggests that increased separation between crystallites is more likely than changes in intermolecular distance. At larger strain values, irreversible characteristics begin to be observed (Figure 3b), which we attribute to the onset of cracking. The irreversible behavior associated with crack formation allows the devices to be “programmed” to have strain-independent properties within a chosen strain range (Figure 3d), providing a method to create reliable stretchable transistors consisting of conventional organic semiconductors. Despite the formation of cracks, the change in mobility shows the same trend in the perpendicular and parallel stretching directions. This is contrary to the findings of O'Connor et al.,39 who found that the mobility increases in the perpendicular direction and decreases in the parallel stretching direction. However, it is consistent with some studies on strained pentacene devices.45 Additional considerations include the role of the electrodes and dielectric. While the conductivity of the S/D electrodes decreased with strain (Figure S10), this is not expected to significantly affect the device performance due to the small currents extracted from the device. The strain-induced changes in dielectric capacitance were consistent with what would be expected based on a Poisson ratio of 0.5 (Figure S11). While increasing capacitance should improve device performance, the effect was evidently overshadowed by changes in the semiconductor layer. The cycling stability of the electrical characteristics was investigated by repeatedly applying 40% strain perpendicular to the charge transport direction. The device characteristics were measured in the unstretched state ∼5 min after completing 1, 10, and 100 cycles, and the transfer curves are depicted in Figure 4a. IOFF remained constant throughout the cycling measurements. After the first cycle (initial programming), ION decreased by 17% after cycle 10 and 28% after cycle 100. This cycling performance could be related to the changes in P3HT morphology and the viscoelastic properties of the substrate. R decreased from ∼1.3 after one cycle to ∼1.2 after 300 cycles, indicating some reorganization of the P3HT (Figure 4b). Observations from AFM (Figure S12) provide support for the morphology change with increased cycle number. In addition to the changes in semiconductor morphology, the viscoelastic properties of the substrate can affect cycling results. Physical crosslinking elastomers, such as the thermoplastic PU used in this work, are known to exhibit large mechanical hysteresis,47 and the stress-strain behavior of the substrate (Figure S13) indicated that viscous deformation increases with increasing cycles. The extension ratio (λ) at zero stress was taken as a measure of the extent of viscous deformation. A plot of 1/λ vs strain displays a similar trend as the normalized ION (Figure 4c), providing support for a relationship between the substrate viscosity and the cycling performance. To further characterize the effect of substrate viscosity, transfer characteristics were collected at different times after the completion of 100 cycles to 40% strain (Figure 4d). Over a 40 minute period, ION increased from 0.65 μA to 0.80 μA, which can be attributed to the continued contraction of the substrate toward its initial dimensions. In contrast, pentacene devices operated within the elastic strain range of the substrate have cycling The of a in the electrical properties the of elastomer viscoelastic properties in stretchable devices. crosslinking elastomers, which typically less viscous may be more for compliant electronics. In stretchable transistors have been fabricated that exhibit transistor characteristics to large the semiconductor between two elastomer was found to the deformation and the formation of cracks. Strain-independent characteristics were achieved by microcracks with an initial strain. measurements within the range up to the initial strain were relatively constant (Figure The cycling performance of the device is related to both the change in P3HT morphology and the viscous deformation in the the of elastomers with viscous work is to a of the of the electrical properties of the devices. While the described devices provide an initial to several in stretchable there are many potential for the dielectric with a more elastomer that solvents could improve the device characteristics. Furthermore, the stretching performance could be improved by a more elastic for applications, the liquid electrodes must be with a state solvents were from and used as P3HT = to = and polyurethane were by and respectively. substrates were by PU in at a of The solution was into a and the was at The substrate thickness was in at a of were for min at The resulting was at for min to and and the top of the was for were in UV/Ozone for min to The substrates were at on a and a was used to the through a to the source and drain to in the spraycoating process, the channel length was PU in was on the electrodes and the was to while with a The electrodes were transferred onto a μm thick PU substrate for with a layer of were by spincoating a solution of in followed by a in P3HT in was spincoated onto the using a (1) for and (2) with for The resulting P3HT films were transferred onto the electrodes at by applying for 2 The PU dielectric was by spincoating from at the spincoating process, the device was from the to that a of the electrodes remained To the eutectic gallium indium liquid was applied as a top gate using a and The manual process of applying the resulted in variation in the of the and the channel width was measured using optical The ratio was used to measure the dielectric properties of the elastomer with strain were fabricated by spraycoating onto without a and then embedding in PU was spincoated on top from and was used as the top electrode. The sheet resistance of the CNT electrodes was measured using a probe connected to a The resistance of the CNT electrodes was investigated by the film into a linear that been in to include was used to make electrical contact to the of the the PU was used to make contact to the bottom CNT and a was used to make contact to the top electrode. The electrical characteristics were collected using an with a To UV-Vis and optical microscope thin PU substrates were used to facilitate deformation. was applied by stretching the substrate and applying to it to a images were collected using a microscope with a and operated in UV-Vis spectra were collected using a with a that could be relative to the Electrical characteristics of the transistors were collected using a probe station in a nitrogen connected to a The source and drain electrodes were using μm to the of the while contact to the gate was using a measurements were by a operated stretching The strain in the device was measured using measurements were conducted in a nitrogen by using a second stretching Electrical measurements were collected in the released state after completing the of cycles. properties of the PU substrate were collected using an with a 100 cycles were completed at a strain rate of This work is by the of and of This research was by the of and the by the a to our and this provides supporting by the materials are and may be for but are not or support from supporting than should be to the The is not responsible for the or functionality of supporting by the than should be to the for the