Electrical properties of the stretchable TFT

Electrical characteristics of a typical SWCNT-AgNW TFT with a density of SWCNT about 25tubesμm⁻¹ are shown in Fig. 2a,b. All devices were tested in air at room temperature. The SWCNT (with 99+% semiconductive SWCNTs) used in this work has a length range from 300nm to 5μm. Channel length of >20μm is reported to be required to decrease the probability of percolative transport through the metallic nanotubes in order to achieve I_ON/OFF>10⁴ (refs 28, 29). A channel length of 100μm and channel width of 5,000μm are thus selected, but not optimized, in the present work for all device testing. A family of output curves appears to be linear at low fields with gate voltage stepped from −1 to −5V. The specific contact resistance between the semiconducting SWCNT network and AgNW electrodes is calculated to be 0.219±0.043Ωcm² through the transmission line model measurements as detailed in Supplementary Fig. 4 and Supplementary Note 2 (refs 30, 31). The contact resistance between SWCNT and AgNW is higher than that between SWCNT and evaporated Ag electrode (0.02Ωcm²)30,31. The work function of the AgNW is measured to be ~4.90eV by Kelvin probe force microscopy, which is higher than the typical work function of Ag (4.73eV), indicating surface of AgNW partially covered by Ag₂O. This surface oxide might have a dominant effect on the interface characteristics with SWCNTs, and thus increase the contact resistance. However, this contact resistance between AgNW and SWCNTs is still negligibly small compared with the channel resistances because of the large channel length used in this work32. Clear current saturation at high drain voltages due to pinch-off is also evident. Moreover, notable increase of the conductance of the channel with the increasing gate voltage from 0 to −5V in 1V steps manifests the effective gate modulation of the EDC.

Open in a separate window

Figure 2

Electrical properties of a representative stretchable TFT and statistical variation studies in the electrical properties of 36 stretchable TFTs.

(a) Output (I_D–V_D) characteristics of a typical SWCNT-AgNW TFT (L=100μm and W=5,000μm) with V_G from 0 to −5V in 1V steps. (b) Transfer (I_D–V_G) characteristics of the same device with V_D from −1.0 to −3.0 in 1.0 step. Inset, I_D–V_G curve at V_D=−2.0V on a logarithmic scale. (c) Transconductance at V_D=−2.0V as a function of V_G. (d–g) Histograms of TFTs showing the statistical distribution of (d) mobility, (e) unit width normalized transconductance, (f) unit width normalized I_ON and (g) the I_ON/OFF.

From the transfer characteristics shown in Fig. 2b, one typical device exhibits a peak I_ON value of 140μA and peak transconductance (g_m) of 54.4μS, corresponding to 28.0μA mm⁻¹ (I_ON/W) and 10.9μS mm⁻¹ (g_m/W) as normalized by the channel width at V_D=−2.0V. Well-defined ON-state and OFF-state currents were observed with a I_ON/OFF of 7,000 at V_D=−2.0V. A gate-sweep-rate-dependent hysteresis was also observed for the SWCNT-AgNW TFT devices (see Supplementary Fig. 5), which is typical for SWCNT-based devices (see Supplementary Note 3)33,34. The mobility of the device is extracted using the following equation: μ=(dI_D/dV_G)(L/V_DC_iW), where μ is the device mobility, I_D the drain current, V_G gate voltage, L channel length, W channel width and C_i specific capacitance of the elastomeric dielectric. C_i was measured from parallel-plate test structures with an average value of ~16.7nFcm⁻² at 12Hz with thickness being ~1.1μm. An unstretched device mobility of 32.5cm²V⁻¹s⁻¹ is thus calculated from the transfer characteristics in Fig. 2b. However, SWCNTs are sparsely distributed in the channel (Fig. 1e), and the parallel-plate capacitor model overestimates the gate capacitance and underestimates the mobility. Using the cylindrical model that takes into consideration of the effect of electrostatic coupling between nanotubes to estimate the gate capacitance19,35,36, a higher mobility of 51.6cm²V⁻¹s⁻¹ is obtained (see Supplementary Notes 4 and 5).

In 7 different batches for a total of 36 TFT devices fabricated, the yield was 85.7%. The I_ON/OFF, peak I_ON value, g_m and mobility were measured in the ranges of 4,500–12,300, 8.6–23.0μA, 32.3–72.1μS and 19.3–43.1cm²V⁻¹s⁻¹, respectively. Fig. 2d–f present the histograms of the statistical variations for these 36 devices in I_ON/OFF, I_ON/W, g_m/W and mobility, respectively. The I_ON/W and g_m/W were measured at V_D=−2.0V, and the mobility was also extracted as described above at V_D=−2.0V. The average of I_ON/OFF ratio, I_ON/W and g_m/W and mobility is 7,200±1,200, 30.0±8.93μA mm⁻¹, 11.2±3.40μS mm⁻¹ and 27.0±7.44cm²V⁻¹s⁻¹, respectively. The variation in the device performance has been widely observed in printed SWCNT transistors11 and could be improved by delicate control of the printing parameters37.

The device characteristics strongly depended on the density of the SWCNTs (see Supplementary Figs 6 and 7), which was controlled by the amount of SWCNT ink drop-casted on the substrates (Fig. 1e). Supplementary Figure 8 shows I_ON/W, g_m/W, I_ON/OFF and mobility as a function of density of nanotubes for V_D=−2.0V. As the density of nanotubes decreases from 25 to 6tubesμm⁻¹, the average of I_ON/OFF increases from 7,200 to 18,300. This increase of I_ON/OFF ratio at low nanotube density is possible due to the decrease in the probability of percolative transport through the metallic nanotubes29,38. As long as the percolation threshold is smaller than the density of CNTs but larger than that of the metallic tubes, all percolation paths from source to drain must involve at least one semiconducting segment. Gate modulation of this ‘week' connection of semiconducting SWCNTs ensures high I_ON/OFF ratio39,40. At high SWCNT densities, the density of metallic SWCNTs can exceed percolation threshold and the network forms continuous paths across the device, which results in high off current (I_OFF) and low I_ON/OFF. As a trade-off, low density of nanotubes leads to a decrease of I_ON/W and g_m/W. As a result of the direct dependence of mobility with g_m, the average mobility also decreases from 27.0 to 9.22cm²V⁻¹s⁻¹ (parallel model) with the decrease of the density of SWCNT from 25 to 6tubes μm⁻¹ (see Supplementary Note 5). In the present work, the SWCNT density for the TFT device was optimized to ~25tubes μm⁻¹ to have a high average mobility of 27.0cm²V⁻¹s⁻¹, and an average I_ON/OFF of 7,200, which meets the demand of an I_ON/OFF ratio of at least ~10³–10⁴ for the application of digital logic circuits28,39.

To investigate the electrical properties of the device under mechanical deformation, the transfer characteristics were further measured during the application of a uniaxial strain along either the channel length (Fig. 3a,b) or channel width (Fig. 3c,d). The I_ON value is declined by 52 and 51% with the strain increase from 0 to 50% along the channel length and width directions respectively (see Supplementary Fig. 9). In the meanwhile, I_OFF also exhibits a slight decrease under strain of up to 50%: the value is dropped by 17 and 46% with respect to the unstretched values along the channel length and width directions, respectively. The decrease in the source-drain channel current during stretching is mainly attributed to the increase in the contact resistance between the SWCNTs when stretched (see Supplementary Fig. 10)19,41,42. The mobility determined from the g_m and capacitance data (see Supplementary Fig. 11) exhibits a decrease, from 32.7 and 28.2cm²V⁻¹s⁻¹ before stretching, to 16.2 and 15.0cm²V⁻¹s⁻¹ at 50% strain along channel length and width directions, respectively. It is suggested that the decrease in capacitance of EDC with increasing strain is the main cause leading to the decrease of mobility (see Supplementary Fig. 11). The chain mobility of PEG in the DEC would be restricted when tensile strain is applied on the EDC, which results in the decrease of dielectric constant (see Supplementary Note 6). The elongation of the channel length between the AgNW-PUA source and drain electrodes can be obviously observed from the magnified photographs of TFTs at specific strain applied along channel length direction (Fig. 3e). The device fails when the strain is over 50% because of the tearing of elastomeric dielectric. Furthermore, I_ON/W, g_m/W and I_ON/OFF at strain of 50% along the channel length and width directions, which can be obtained from the transfer characteristics from Fig. 3a–c, are calculated to be 11.0μAmm⁻¹, 8.22μSmm⁻¹ and 8,400 along the channel length direction, and 10.4μAmm⁻¹, 7.61μSmm⁻¹ and 15,400 along the channel width direction, respectively. It is of noteworthy that the electrical performance of the stretchable TFTs at 50% strain is still comparable to most unstretchable TFT devices using SWCNTs as the channel material.

Open in a separate window

Figure 3

The device performance changes with tensile strain.

(a) Typical transfer characteristics (V_D=−2.0V) of a TFT device under specific tensile strain applied along the channel length direction. The insets show log-scale characteristics. (b) I_ON, I_OFF and mobility as a function of applied strain along the channel length direction. (c) Typical transfer characteristics (V_D=−2.0V) of a device under specific tensile strain applied along the channel width direction. The insets show log-scale characteristics. (d) I_ON, I_OFF and mobility as a function of applied strain along the channel width direction. (e) Magnified photographs of a device at specified strains applied along the channel length direction. Scale bar=5mm.

A fatigue test was subsequently performed by subjecting the TFT to 500 cycles of continuous stretch–relaxation between 0 and 20% strains along channel length or width direction. Figure 4 shows the electrical transfer characteristics, I_ON and I_OFF change during the continuous stretch–relaxation cycles. The I_ON for both strain direction decreased lightly over the 500 stretch–relaxation cycles, and the I_ON value remained fairly high value of ~60μA (initial value ~140μA) along channel length direction and ~90μA (initial value ~160μA) along channel width direction after 500 cycles (see Supplementary Fig. 9). In contrast, the I_OFF values remain relative constant throughout the cycling measurements along either direction. The calculated I_ON/OFF after 500 stretching–relaxing cycles along the channel length and width directions is ~10,000 and 13,000, respectively. The mobility degraded relatively fast in the initial 200 cycles for both direction, but slowed down in the following 300 cycles. Mobility of 11.7 and 17.6cm²V⁻¹s⁻¹ were obtained from the transfer characteristics curves after 500 stretch–relaxation cycles along channel length and width direction respectively. The retention of the device mobility after repeated strain cycles is still much higher than data reported in literature for all-printed SWCNT-based transistors37.

Open in a separate window

Figure 4

Fatigue testing when stretching and releasing 500 times.

(a) Typical transfer characteristics (V_D=−2.0V) of a TFT device after specified cycles of 20% tensile strain applied along the channel length direction. The insets show log-scale characteristics. (b) Plots of I_ON and I_OFF at 0% strain during 500 cycles of continuous stretching–relaxing between 0 and 20% strains along the channel length direction. (c) Typical transfer characteristics (V_D=−2.0V) of a device after specified cycles of 20% tensile strain applied along the channel width direction. The insets show log-scale characteristics. (d) Plots of I_ON and I_OFF at 0% strain during 500 cycles of continuous stretching–relaxing between 0 and 20% strains along the channel width direction.

We speculate this good device performance for the device after subjecting repeated stretch–relaxation cycles is resulted from the unique structure where the EDC infiltrate into the SWCNT network and wraps around the nanotubes (Fig. 1a). The elastomeric dielectric wrapping around the SWCNT should constrain the junction disjoining and tube sliding, help distribute the stress more uniformly under tensile strain and thus increase the stretchability and stability of the SWCNT network under large strains6,7,43,44. Figure 5a,b display the normalized resistance (R/R₀, the ratio of the instantaneous resistance at a specific tensile strain to the initial resistance at zero strain) for the SWCNT coated on PUA substrate with and without the EDC overcoat during 1,000 cycles of tensile stretching-releasing between 0 and 20% strains. It can be seen that after a gradually increase in the initial 100 cycles, the resistance of the SWCNT Network with EDC overcoat becomes very stable in the subsequent 900 cycles. The resistance increase during the first 100 cycles can be attributed to the alignment and sliding of the SWCNT network during deformation and the viscoelasticity of the elastomeric dielectric6,44. Overall, the elastomeric dielectric helps preserve the SWCNT network during large deformation very well. This trend of resistance change for the SWCNT network is consistent with the trend of mobility change for the TFT during continuous stretching–relaxing cycles as shown in Fig. 4b,d. In contrast, it can be observed from Fig. 5b that the SWCNT network without EDC overcoat exhibits a gradual increase in resistant continuously over 1,000 cycles.

Open in a separate window

Figure 5

Mechanical compliance of SWCNT network on PUA substrate.

Normalized transient resistance of SWCNT coated on PUA substrate with (a) and without (b) the elastomeric dielectric overcoat during 1,000 cycles of tensile stretching and releasing between 0 and 20% strains.

Preparation of AgNW-PUA composite source/drain and gate electrode

To pattern the source/drain on glass, a dispersion of AgNWs in methanol (0.2wt%) was spray-printed on glass through a patterned shadow mask using Paasche air brush. To fabricate the gate electrode on glass, a dispersion of AgNW (0.5wt%) in IPA was drop-casted into a thin line at one end of the substrate and drawn down with a Meyer rod to spread the solution to form a uniform coating. The resulting transparent conductive coating (both source/drain and gate) on the glass substrates was then coated with a precursor solution consisting of 100 weight parts CN990, 20 parts SR540 and 1 part photoinitiator. The coatings were cured under UV at 2.5Wcm^-2 intensity and peeled off of the release substrate. The AgNW source/drain electrode with channel length of 100μm and channel width of 5,000μm and gate electrode were transferred as the conductive surface of the resulting transparent composite electrodes. Both the AgNW-PUA composite source/drain and gate electrode were prepared with sheet resistance controlled to be 15Ωsq⁻¹,8.

Preparation of elastomeric dielectric

The precursor of DEC was formulated by mixing the components CN990, SR259, SR550 and photoinitiator, in an optimized weight ratio of 1:1:1:0.015. CN990 was chosen for the high stretchability and good bonding of its homopolymer with the AgNW-PUA source/drain and gate composite electrode. SR259 and SR550 were chosen for the good flexibility and high dielectric constant of their homopolymer and low viscosity of their monomer (25 and 19c.p.s. @ 25^oC, respectively). SR550 can serve as a polar side chain to further improve the dielectric constant.

Fabrication of fully transparent and stretchable transistor

The device fabrication started with printing SWCNT on the AgNW-PUA composite source/drain electrode. First, SWCNT ink was prepared by mixing 99%+ Semiconducting -SWCNT aqueous solutions (used as-received from Nanointegris Inc.), DI-water, propylene glycol and Fluorosurfactant at the ratio of 25:5:5:1, followed by untrasonicating for 20min. Subsequently, a certain amount of SWCNT ink was uniformly drop-casted on the conductive surface of AgNW-PUA source/drain composite electrode, followed by heating up to 90^oC on a hot plate to dry the SWCNT network. SWCNT densities on the source/drain channel was controlled by the amount of SWCNT ink being 6, 15 and 30μl on 0.5 × 1.5cm² AgNW-PUA composite source/drain electrode to obtain SWCNT densities of ~6, ~13 and ~25tubesμm⁻¹, respectively. The SWCNT/AgNW-PUA source/drain was rinsed with Deionized water and IPA alternatively to remove the sodium dodecyl sulphate surfactant and Fluorosurfactant, followed by drying in vacuum for 1h. SWCNT was patterned by O₂ plasma etching to the source/drain electrode. Then, the elastomeric dielectric precursor was overcoated on the SWCNT/AgNW-PUA source/drain substrate, spin-printed at 8,000r.p.m. for 1min and then UV-cured in nitrogen gas. Transparent tape (3M) was used to define the openings for the source/drain electrode and peeled off after polymerizing the elastomeric dielectric. Finally, the resulting elastomeric dielectric/SWCNT/AgNW-PUA source/drain structure was laminated with another 15Ωsq⁻¹ AgNW-PUA composite electrode as gate on top of the dielectric layer to complete the device fabrication.

The work reported here was supported by the Air Force Office of Scientific Research (Program Director: Dr Charles Lee; Grant no. FA9550-12-1-0074).