Research On Complementary Integrated Circuits Based On N-IGZO And P-SnO Thin-Film Transistors And Device Modeling

At present,thin-film transistors(TFTs)have developed into a huge industry,mainly used in the field of displays.The mainstream industrial TFT technologies are low-temperature polysilicon(LTPS),amorphous silicon(a-Si),and oxide semiconductors.Other new types of TFTs like organic semiconductors and carbon nano tubes developed rapidly,but their technological maturity still need to be further developed to solve the stability,purity,etc.With the emergence of new electronic technologies,represented by wearable electronics and flexible displays,thin film integrated circuits have become increasingly important,and new requirements have been put forward for the performance of thin-film materials including low cost,flexibility,and high stability.a-Si has disadvantages such as low field-effect mobility and threshold voltage drift under visible light or bias.LTPS shows low homogeneity limited by polycrystalline characteristics and its cost is high due to the complex process.Recent years,oragnic semiconductors,carbon nano tubes,etc.developed rapidly.However,due to e.g.,stability and purity issues,their industrial TFT technologies need to be further improved.The emergence of amorphous indium gallium zinc oxide(1GZO)makes oxide semiconductors attract more and more attention.Compared to other mainstream TFT technologies,oxide semiconductors are very promising materials with the advantages of transparency under visible light,high field-effect mobility,good stability and homogeneity,low-temperature and even room-temperature process(so as to deposite on flexible substrates),and low cost.Most of the reported oxide-based circuits are unipolar,especially those based on 1GZO TFTs,because of the high challenge of obtaining high-performance P-type oxide TFTs.However,compared with unipolar technology,complementary technology shows advantages in all aspects,including but not limited to low power consumption,strong anti-interference ability,high integration,and rail-to-rail output.In order to realize oxide-based complementary circuits,some research groups adopted hybrid complementary technology,by combining N-type oxides with P-type organics or other materials,which achieved high circuit performances.However,hybrid complementary technology brings new challenges of process compatibility and complexity,whcih result in high cost.Among the reported limited numbers of P-type oxide materials,tin oxide(SnO)has been regarded as the most promising one,mainly due to high hole mobility,low-temperature large-scale fabricated,and good stability.Complementary technology based on all-oxide semiconductors is the best solution theoretically.Although several circuits based on complementary oxide semiconductors are reported,the development is still far behind in comparison to unipolar circuits based on N-type oxides.The reported circuits limited to simple circuit units with single function and low integration degree such as inverters,logic gates,and ring oscillators.More importantly,in terms of sequential logic circuits,the research on flip-flops is still lacking,and there has been no related publication.In order to accelerate the applications of oxide semiconductors in emerging fields,it is imperative to improve the functions and scale of all-oxide complementary circuits.The development of integrated circuit technology requires accurate device models to predict accurately circuit behaviors,which can make circuit design more efficient and cost-effective.The mainstream modeling method is semi-empirical method,which adds some fitting functions or empirical parameters on the basis of carrier transport mechanism and material properties.The reported oxide TFTs models are limited basically to N-type oxide TFTs,which meets the requirement of unipolar oxide circuit simulation.However,due to that the physical mechanisms of P-type oxide TFTs are not yet clear,using semi-empirical method is challenged,and it is a common problem that other new semiconductors are facing or about to face.Therefore,looking for a modeling method without physical mechanism and material characteristics is a key to promoting the application of simulation technology on all-oxide complementary circuits,and also can act as a reference for other new semiconductors.In view of the limited but timely development of oxide semiconductors in complementary circuits,based on N-type IGZO TFTs and P-type SnO TFTs,this thesis work has achieved high-performance complementary inverters with large-area uniformity and good stability.Base on this,a complementary static random access memory(SRAM)and three types of flip-flops,especially the D-type edge-triggered flip-flop(DFF),have been fabricated for the first time.Based on the DFFs,a 2-bit reversible counter with 90 TFTs has been designed and successfully developed,demonstrating the possibility of large-scale complementary integration of N-IGZO and P-SnO in terms of function and yield.The electrical performance of these circuits are analyzed in detail and compared with the values of the reported circuits in literature.The artificial neural network(ANN)is applied to SnO TFT modeling for the first time,and output characteristic simulation is realized using circuit simulation software.The main contents of this thesis are as follows:1.High-performance invertersExcellent device uniformity is a prerequisite for normal operation of large-scale integrated circuits.Because that P-type oxide TFTs are still in early stage,and few study have reported the dive uniformity.In this thesis work,the field effect mobility and current switching ratio of SnO TFTs in the linear region are 0.7 cm2/Vs and 2.6 x 104,respectively.Eight SnO TFTs have been selected randomly,and then the average and standard deviation of their threshold voltages under different VDS were evaluated.These SnO TFTs exhibit extremely excellent uniformity,because the transfer characteristic curves almost completely conincide.The inverters are the cornerstpne of modern circuit systems.There are three key parameters to evaluate inverter performance:switching threshold voltage(VSP),voltage gain,and noise margin level(transition width).When the ratio(N)of the width-to-length ratio of SnO TFT to the width-to-length ratio of IGZO TFT is 8,the VSP closes to the ideal value,which is equal to VDD/2.Twelve inverters with N=8 and different width-to-length ratios of IGZO TFT have been randomly selected in the working area of 1 cm × 1 cm to analyize the dispersion,and VSP is 4±0.022 V when VDD=8 V,which means that the VSP of each inverter reaches a quite ideal value,and the high and low noise margin are balanced.The transition width is only 1.04 ± 0.024 V,and the effective input voltage ratio is as high as 87.5%,indicating that the anti-interference ability is extremely strong.The voltage gain is 113±16.5 with a maximum value of 142,breaking the record of the highest voltage gain of the reported oxide-based complementary inverters.In comparison to the performance shortages of the reported oxide-based complementary inverters,our fabriciated inverters achieved ideal VSP,high and balanced noise margin level,and a record high voltage gain.In addition,such high performance is repeatable which is rather amazing for the future mass production..2.High-performance SRAMThe SRAM cell is an essential circuit module for data processing.The all-oxide complementary SRAM cell was fabricated for the first time,with an area of only 0.0208 mm2,which is the smallest value among the reported SRAM cells based on flexible semiconductors.Based on the static voltage transfer characteristic(SVTC)curve,the stabilities during the read,write and hold state were analyzed,and the corresponding noise margins were extracted graphically,which are 1.43,1.67 and 2.3 V,respectively.In addition,fatigue tests and ambient air stability tests(for 5 months)were performed on the SRAM cell.These results show that the SRAM cell has good stability while working or being exposed to the ambient air for a long time.Since SRAM cells are more likely to fail during read operation,the N-curve method was used to analyze read stability,with static current and voltage noise margins of 13 ?A and 2.05 V,respectively.Writing time is also a very important parameter and should be as short as possible.It can be extracted from the dynamic waveforms.The transition time of writing "1" and"0" is 121 and 82 ?s,respectively,which is the fastest write speed among the flexible semiconductors based SRAM cells.The results show that the fabricated oxide-based complementary SRAM cell has excellent performance,indicating high potential for data storage and processing in large-scale flexible electronics.3.Sequential logic circuitsAt present,there are no reports on oxide-based complementary DFFs,let alone sequential logic circuits with complex functions and higher integration scale.The research progress is far behind unipolar oxide based circuits.Two common flip-flops,DFF and master-slave JK FF,were fabricated for the first time based on complementary oxides.The robustness of the output signal of the DFF was analyzed.E,ven if the voltage difference between high and low level of the input signal is as small as 1 V(high level voltage is 4.5 V,and low level voltage is 3.5 V),the output signal can still remain unchanged.The delay time of the DFF determines the supported maximum clock frequency,which can be also extracted from the dynamic waveforms.The propagation delay times from "0" to "1" and from "1" to "0" are 17 and 40 ?s,respectively,and the corresponding transmission delay times are 31.6 and 46.3 ?s,respectively.Compared with reported organic-based complementary DFFs,the delay time is very short.Based on the high-performance DFFs,a 2-bit binary reversible counter using 90 TFTs to realize the functions of add count and subtract count was designed and fabricated successfully.The yield of the circuit is 55%,which means that the yield of single TFT is 99.34%.The results demonstrate the high possibility of large-scale complementary integration using N-IGZO and P-SnO TFTs from the point views of function and yield.4.SnO TFT device modelSimulation technology is an indispensable part during the development of thin-film integrated circuit technology.However,due to the lack of accurate SnO TFT models,the simulation of all-oxide complementary circuits is very difficult.Artificial neural network was applied to SnO TFT modeling for the first time.Multi-layer perceptron neural network and reverse error propagation algorithm were adopted.The model has three input variables,namely width-to-length ratio(W/L),drain-source voltage(VDS)and gate-source voltage(VGS),and the output variable is drain-source current(IDS).The training samples were output characteristic curves of SnO TFTs with different WIL values.The influence of different sample size on performance and iteration time was discussed.The appropriate sampling intervals of VDs and VGS were-0.8 and-0.18 V,respectively.In addition,the number of neurons and iterations were set to 50 and 500,respectively.In order to evaluate the accuracy of prediction,the indicator of mean absolute relative error(MARE)was introduced.A part of the-voltage range of VDS was used to train the neural network,and the remaining voltage range was used to predict and verify the IDS.The on-state current MARE was 2.25%.The same method was applied to VGS,and the on-state current MARE was 0.3%.It is obvious that the established SnO TFT model has high accuracy and can meet the requirements of circuit simulation.Finally,the ANN model was converted to Pspice model for circuit simulation.