Bendable Single Crystal Silicon Nanomembrane Thin Film Transistors with Improved Low-Temperature Processed Metal/n-Si Ohmic Contact by Inserting TiO2 Interlayer

Bendable single crystal silicon nanomembrane thin film transistors (SiNMs TFTs), employing a simple method which can improve the metal/n-Silicon (Si) contact characteristics by inserting the titanium dioxide (TiO2) interlayer deposited by atomic layer deposition (ALD) at a low temperature (90 °C), are fabricated on ITO/PET flexible substrates. Current-voltage characteristics of titanium (Ti)/insertion layer (IL)/n-Si structures demonstrates that they are typically ohmic contacts. X-ray photoelectron spectroscopy (XPS) results determines that TiO2 is oxygen-vacancies rich, which may dope TiO2 and contribute to a lower resistance. By inserting TiO2 between Ti and n-Si, Ids of bendable single crystal SiNMs TFTs increases 3–10 times than those without the TiO2 insertion layer. The fabricated bendable devices show superior flexible properties. The TFTs, whose electrical properties keeps almost unchanged in 800 cycles bending with a bending radius of 0.75 cm, obtains the durability in bending test. All of the results confirm that it is a promising method to insert the TiO2 interlayer for improving the Metal/n-Si ohmic contact in fabrication of bendable single crystal SiNMs TFTs.


Introduction
Flexible electronics is an important development direction in the field of future electronics. Scientists can use flexible materials to fabricate advanced electronic devices, such as transistor arrays for optional folding and stretching, bendable flexible screens, or some sensors which can be integrated on the clothing [1][2][3][4][5][6][7][8][9]. Advances in various flexible electronic technologies, including solar cells, sensors and displays, have been driven by the use of flexible organic materials. However, organic-based semiconductors suffer from poor device performance due to their low carrier mobility and their chemical/thermal instability. Recently, the discovery of single-crystal Si nanomembranes (SiNMs) has fascinated the flexible electronics community because of their high carrier mobility, stable chemical/thermal properties and flexibility. Particularly, SiNMs released from silicon-on-insulator (SOI) become one of the best choices owing to their outstanding electrical properties, mature fabrication techniques and commercial feasibility at relatively lower cost [10][11][12][13][14].
For SiNMs, effectively doping them to form effective ohmic contacts and realizing low contact resistivity are both critical in realizing high speed operation. Nevertheless, many flexible substrates are soft and have very low processing temperature tolerance. For example, ITO/PET substrates just can withstand the highest temperature as low as 150 • C [11]. Hence, the traditional high-temperature processing cannot be directly used. This challenge has been partially overcome by employing a pre-doped (ion implantation and annealing before SiNMs release) SiNMs transfer and gate-last TFTs fabrication process [10,11]. However, a low-temperature process to achieve a lower Metal/Si Ohmic contact is still urgently pursued.
On the basis of the works above, we reports a simple method that using ALD technology deposits TiO 2 at a low temperature of 90 • C to further improve the contact between the source/drain regions and metal electrodes. Current-voltage characteristics of Ti/insertion layer (IL)/n-Si structures, XPS results of TiO 2 and the normalized current-voltage characteristics of bendable single crystal silicon TFTs with different cycles of TiO 2 are obtained and described in detail. By inserting titanium dioxide, good ohmic contacts are formed between the source/drain regions and metal electrodes. I ds of bendable single crystal SiNMs TFTs increases 3-10 times than those without the TiO 2 insertion layer. The TFTs, whose electrical properties keeps almost unchanged in 800 cycles bending with a bending radius of 0.75 cm, obtains the durability in bending test.  Figure 1a shows structure schematic of a bendable single crystal silicon TFT without inserting TiO 2 and Figure 1b shows structure schematic of a bendable single crystal silicon TFT with inserting different cycles of TiO 2 . Figure 1c shows that the devices fabrication process was started with silicon-on-insulator wafer (SOI) (Soitec by Smartcut with 200 nm top Si which is doped boron whose level is 1 × 10 14 cm −3 and 200 nm buried oxide). In our process of making n-channel TFTs, the source/drain regions were first formed on the SOI substrate via phosphorus ion implantation with a dose of 5 × 10 15 cm −2 and an energy of 30 keV followed by annealing in RTP at 1000 • C for 20 s in N 2 ambient to activate the implanted dopants. The doping concentration of the contact area is~10 19 cm −3 . SOI wafer was patterned by lithography to form the hole patterns. RIE was used to etch the holes. 33% hydrofluoric acid (HF) etched buried oxide (SiO 2 ) through a lot of etching holes above SiO 2 [15][16][17][18][19]. Completely etching buried oxide took about 48 h so that the top Si dropped on the bottom silicon substrate by Van der Waals force [20][21][22]. Deionized water was used to wash away residual HF to be ready for transfer. In order to avoid being dislocated and scattered of the SiNMs which was immersed in HF, we designed the mask including a lot of 1 × 1 cm units so that the SiNMs formed a very large area. Even if the buried oxygen was completely etched off, there was no displacement and scatter of the SiNMs on the bottom Si. At the same time, this design increased the transfer area reaching 1 cm 2 . A flat piece of polydimethylsiloxane (PDMS) was brought into conformal contact with the top surface of the wafer and then rapidly peeled back to pick up SiNMs. The interaction between SiNMs and PDMS is sufficient to pick up SiNMs with good efficiency, nearly 100%. A flexible ITO/PET substrate served as the target substrate. The target substrate was washed with acetone and alcohol, rinsed with deionized water and then dried with a stream of nitrogen. Treating the ITO/PET substrate with a short O 2 plasma (20 sccm O 2 flow with 50 W rf power for 10 s) promoted adhesion between it and a spin coating dielectric layer of epoxy (4000 rpm for 30 s of SU8-2002). Then the SiNMs on PDMS was brought against the epoxy layer (SU8-2002). SiNMs was transferred to the ITO/PET substrate by gently pressing and slowly peeling up PDMS. The epoxy layer was cured at 100 • C for 1 min, exposed to UV light from the backside of the sample for 10 s and finally post-baked at 100 • C for 1 min. Photolithography defined a pattern on the substrate. RIE etched the Si to form MESA. O 2 plasma was used to remove photoresist. After 5% HF solution immersion for 2 min and de-ionized water rinse for 2 min, N 2 gun blew dry the samples. Then the samples were immediately loaded into PicosunTM R-200 Advanced ALD chamber. Titanium tetrakis (dimethylamide) (TDMATi) was used as Ti source and H 2 O was used as oxygen source. During the process, TDMATi source bottle was heated to 120 • C and the N 2 carrier flow was set to 15 standard cubic centimeter per minute (sccm), pulse time and purge time were 100 ms and 40 s respectively. For oxygen source, the N 2 carrier flow was set to 15 sccm, pulse time was 100 ms and purge time was 40 s. The whole deposition process was carried out at 90 • C. There were two different cycles of 5 and 10. Using these two conditions formed~0.5 and~1 nm TiO 2 on the SiNMs, respectively. Photolithography defined a pattern on the samples. Put them into a PRO LINE PVD 75 SYSTEM (Kurt J. Lesker Company, Pittsburgh, PA, USA) to deposit Ti (100 nm) by electron beam evaporation. Finally, the source and drain metal contacts formed on the low-resistive source and drain regions followed by lift-off without any further thermal treatment. R-200 Advanced ALD chamber. Titanium tetrakis (dimethylamide) (TDMATi) was used as Ti source and H2O was used as oxygen source. During the process, TDMATi source bottle was heated to 120 °C and the N2 carrier flow was set to 15 standard cubic centimeter per minute (sccm), pulse time and purge time were 100 ms and 40 s respectively. For oxygen source, the N2 carrier flow was set to 15 sccm, pulse time was 100 ms and purge time was 40 s. The whole deposition process was carried out at 90 °C. There were two different cycles of 5 and 10. Using these two conditions formed ~0.5 and ~1 nm TiO2 on the SiNMs, respectively. Photolithography defined a pattern on the samples. Put them into a PRO LINE PVD 75 SYSTEM (Kurt J. Lesker Company, Pittsburgh, PA, USA) to deposit Ti (100 nm) by electron beam evaporation. Finally, the source and drain metal contacts formed on the lowresistive source and drain regions followed by lift-off without any further thermal treatment.

Device Characterization
Etching depth of SiNMs was measured by Stylus Profiler (Bruker Dektak XT, Bremen, Germany). The XPS testing of TiO2 samples was performed on Thermo escalab 250Xi (Thermo Fisher Scientific , Waltham, MA, USA) using monochromatic Al-Ka (1486.6 eV) as the radiation source. The I-V characteristics of Ti/insertion layer (IL)/n-Si metal-insulator-semiconductor (MIS) structures and bendable single crystal silicon TFTs were both measured by using Keithley 1500 semiconductor characterization system (Tektronix, Inc., Beaverton, OR, USA). All the measurements were performed under ambient atmosphere at room temperature without encapsulation.  Figure 2b shows a high magnification optical images of the MIS structure whose Ti pad diameter is 300 μm. Figure 2c shows the I-V characteristics of Ti/insertion layer (IL)/n-Si (doping level is ~10 19 cm −3 ) MIS structure with different ALD cycles. It is obvious that the characteristic curve of devices without TiO2 (0 cycle) is curving which is a typical Schottky contact. It indicates that even if the source/drain area is heavily

Device Characterization
Etching depth of SiNMs was measured by Stylus Profiler (Bruker Dektak XT, Bremen, Germany). The XPS testing of TiO 2 samples was performed on Thermo escalab 250Xi (Thermo Fisher Scientific, Waltham, MA, USA) using monochromatic Al-Ka (1486.6 eV) as the radiation source. The I-V characteristics of Ti/insertion layer (IL)/n-Si metal-insulator-semiconductor (MIS) structures and bendable single crystal silicon TFTs were both measured by using Keithley 1500 semiconductor characterization system (Tektronix, Inc., Beaverton, OR, USA). All the measurements were performed under ambient atmosphere at room temperature without encapsulation.  Figure 2b shows a high magnification optical images of the MIS structure whose Ti pad diameter is 300 µm. Figure 2c shows the I-V characteristics of Ti/insertion layer (IL)/n-Si (doping level is~10 19 cm −3 ) MIS structure with different ALD cycles. It is obvious that the characteristic curve of devices without TiO 2 (0 cycle) is curving which is a typical Schottky contact. It indicates that even if the source/drain area is heavily doped, Ti/n-Si (doping level is~10 19 cm −3 ) will not form a good ohmic contact without any treatment. Amazingly, the characteristic curves of devices which are deposited by the TiO 2 insertion layer are all typically good ohmic contacts. The contact resistance of 5 cycle is the smallest and the contact resistance of 10 cycle is the second smallest. However, with the increase of the cycle (the thickness of TiO 2 ), the contact resistance becomes greater and greater. The explanation for this is that the thickness of titanium dioxide exceeds a certain value leading to carriers passing through the insertion layer at lower tunneling rate [23,24]. This will reduce the current, with presenting a larger contact resistance. Therefore, we chose 5 and 10 cycles of titanium dioxide inserting into the TFTs to compare those without titanium dioxide insertion layer. lower tunneling rate [23,24]. This will reduce the current, with presenting a larger contact resistance. Therefore, we chose 5 and 10 cycles of titanium dioxide inserting into the TFTs to compare those without titanium dioxide insertion layer.   Figure 3b shows that O 1 s results with thin TiO2, shoulder left to O 1 s peak is obvious. This shoulder located at 531.5 eV represents for oxygen vacancies and it indicates that thin sample has more oxygen vacancies. The illustration in the upper left corner shows the fitting curve of 300 cycles. It can be seen that the peak of oxygen vacancies is not obvious. In conclusion, there are some oxygen vacancies in thin TiO2 film, which will dope TiO2 and make it more conductive [24]. Due to the presence of a donor band related to oxygen vacancies which can provide more electrons, TiO2 behaves as an n-type semiconductor and exhibits good electrical conductivity.   Figure 3b shows that O 1 s results with thin TiO 2 , shoulder left to O 1 s peak is obvious. This shoulder located at 531.5 eV represents for oxygen vacancies and it indicates that thin sample has more oxygen vacancies. The illustration in the upper left corner shows the fitting curve of 300 cycles. It can be seen that the peak of oxygen vacancies is not obvious. In conclusion, there are some oxygen vacancies in thin TiO 2 film, which will dope TiO 2 and make it more conductive [24]. Due to the presence of a donor band related to oxygen vacancies which can provide more electrons, TiO 2 behaves as an n-type semiconductor and exhibits good electrical conductivity.

Results and Discussion
vacancies and it indicates that thin sample has more oxygen vacancies. The illustration in the upper left corner shows the fitting curve of 300 cycles. It can be seen that the peak of oxygen vacancies is not obvious. In conclusion, there are some oxygen vacancies in thin TiO2 film, which will dope TiO2 and make it more conductive [24]. Due to the presence of a donor band related to oxygen vacancies which can provide more electrons, TiO2 behaves as an n-type semiconductor and exhibits good electrical conductivity.   Figure 4d shows the I ds of 0, 5, 10 cycle at V gs = 5 v. It presents that the current of 10 cycles is three times the 0 cycle one. The current of 5 cycles is ten times the 0 cycle one. These results are consistent with results of Figure 2c. The reason is that inserting the titanium dioxide of appropriate thickness between Ti and n-Si can effectively restrain the Fermi energy level pinning effect of n-type silicon so as to improve the Ti/n-Si ohmic contact, reduce the contact resistance and increase the current driving ability.  Figure  4d shows the Ids of 0, 5, 10 cycle at Vgs = 5 v. It presents that the current of 10 cycles is three times the 0 cycle one. The current of 5 cycles is ten times the 0 cycle one. These results are consistent with results of Figure 2c. The reason is that inserting the titanium dioxide of appropriate thickness between Ti and n-Si can effectively restrain the Fermi energy level pinning effect of n-type silicon so as to improve the Ti/n-Si ohmic contact, reduce the contact resistance and increase the current driving ability.   Figure 5a presents transfer characteristics of TFTs with different cycles of TiO 2 . I on/off of 0 cycle is 10 2 , 5 cycles is 10 4 and 10 cycles is 10 3 . I on of 0 cycle is smallest, 5 cycles is largest and 10 cycles is medium. This result is consistent with results of Figures 2c and 4d. Thin TiO 2 which restrains the Fermi energy level pinning effect of n-Si can improve the interface contact between Ti/n-Si contributing to increase I ds [23][24][25]. I off of 0 cycle is largest, 5 cycles is smallest and 10 cycles is medium. For transistors, the smaller the I off , the better the performance of the device. After analysis, it may be that ALD growth of titanium oxide plays anneal role so that the interface contact between epoxy layer (SU8-2002) and SiNMs is better than that without depositing TiO 2 . It promotes the ability to control current of gate electrode. To test this conjecture, we put the TFT without depositing TiO 2 into ALD annealing at 90 • C for 5 min (the time of depositing 5 cycles TiO 2 ) and then compared the transfer characteristics with unannealed one. Results are as Figure 5b shown, the I off of without TiO 2 but with annealing is an order of magnitude smaller than that without TiO 2 and without annealing, which is consistent with results of Figure 5a. Hence, it indicates that even if annealing at 90 • C can also improve the interface contact between epoxy layer (SU8-2002) and SiNMs with reducing interface defects and enhancing the ability to control current of gate electrode to reduce I off . Thus, the combination of the inserted TiO 2 layer and very low-temperature annealing process improves the metal/Si ohmic contact.    Figure 6b shows the digital photographs of the flexible TFTs. Figure 6c shows an image of one device of TFTs under an optical microscope. Figure 6d shows a bending test of the flexible TFTs with a bending radius of 0.75 cm. Obtaining the durability in bending conditions is essential to wearable applications [26]. The electrical properties of the devices do not change significantly in 800 cycles with a bending radius of 0.75 cm, confirming a reliable performance for flexible operations.   Figure 6b shows the digital photographs of the flexible TFTs. Figure 6c shows an image of one device of TFTs under an optical microscope. Figure 6d shows a bending test of the flexible TFTs with a bending radius of 0.75 cm. Obtaining the durability in bending conditions is essential to wearable applications [26]. The electrical properties of the devices do not change significantly in 800 cycles with a bending radius of 0.75 cm, confirming a reliable performance for flexible operations. Figure 6a is an image of the flexible TFTs fixed on a probe station with the bending radius of 0.75 cm. Figure 6b shows the digital photographs of the flexible TFTs. Figure 6c shows an image of one device of TFTs under an optical microscope. Figure 6d shows a bending test of the flexible TFTs with a bending radius of 0.75 cm. Obtaining the durability in bending conditions is essential to wearable applications [26]. The electrical properties of the devices do not change significantly in 800 cycles with a bending radius of 0.75 cm, confirming a reliable performance for flexible operations.

Conclusions
A simple method was used to improve Ti/n-Si contact characteristics by inserting ALD deposited TiO2. The fabrication temperature of inserting ALD deposited TiO2 is below 120 °C which is the

Conclusions
A simple method was used to improve Ti/n-Si contact characteristics by inserting ALD deposited TiO 2 . The fabrication temperature of inserting ALD deposited TiO 2 is below 120 • C which is the highest temperature plastic can withstand. Hence, we were able to combine this method into the manufacture of flexible thin film transistors. The results show that this method is not only feasible but also effective. It can change the Schottky contact into a satisfactory ohmic contact so that increase current drive capability. The electrical properties of the flexible TFTs do not change significantly with bending 800 cycles, confirming a reliable performance for flexible operations.
Author Contributions: C.Z. conceived the idea, designed the experiment and guided the experiment. J.Z. conducted most of the device fabrication and data collection and wrote the manuscript; C.Z. and Y.Z. revised the manuscript; D.C., W.Z., H.X. helped the device measurement, J.Z. helped the data analysis. Y.H. supervised the group. All authors read and approved the manuscript.