Periodic Reporting for period 1 - BAYFLEX (BAYesian Inference with FLEXible electronics for biomedical Applications)
Reporting period: 2023-04-01 to 2024-03-31
The goal of the BAYFLEX project is to use fault-tolerant stochastic computing with all-organic technologies on flexible substrates to demonstrate an integrated flexible printed circuit that can detect electrophysiological signals and classify them into categories. The project brings together a diverse consortium with expertise in modeling emerging devices, biologically inspired circuit design, experts in machine learning involving electrophysiological data and teams with expertise in OTFT and OECT fabrication.
The work is grouped into three objectives: Objective 1: Model and simulate Develop physical design oriented models of discrete flexible Organic Thin Film Transistors (OTFTs) and complementary Organic Electro-Chemical Transistors (OECTs), simulate circuits for transforming electrophysiological signals into probabilistic bit streams and simulate and design circuits for Bayesian inference using cascaded Muller C-element circuits, Objective 2: Fabricate Develop and realize scalable technologies on flexible substrates for sensing electrophysiological signals, simple signal processing, converting these signals into stochastic streams that can be used for Bayesian inference, Objective 3 Demonstrate and Benchmark: Benchmark the BAYFLEX concept to find suitable electrophysiological tasks, benchmark the technologies to ensure optimal results and assess risks, demonstrate the BAYFLEX concept using the technology from objective 2 on a simple electrophysiological task (TRL 3-4). This objective also includes a life cycle analysis to demonstrate its reduced environmental impact compared to alternatives.
The work is grouped into three objectives: Objective 1: Model and simulate Develop physical design oriented models of discrete flexible Organic Thin Film Transistors (OTFTs) and complementary Organic Electro-Chemical Transistors (OECTs), simulate circuits for transforming electrophysiological signals into probabilistic bit streams and simulate and design circuits for Bayesian inference using cascaded Muller C-element circuits, Objective 2: Fabricate Develop and realize scalable technologies on flexible substrates for sensing electrophysiological signals, simple signal processing, converting these signals into stochastic streams that can be used for Bayesian inference, Objective 3 Demonstrate and Benchmark: Benchmark the BAYFLEX concept to find suitable electrophysiological tasks, benchmark the technologies to ensure optimal results and assess risks, demonstrate the BAYFLEX concept using the technology from objective 2 on a simple electrophysiological task (TRL 3-4). This objective also includes a life cycle analysis to demonstrate its reduced environmental impact compared to alternatives.
Objective 1
The work on the model and simulate objective uses device measurements to develop models and circuits. During RP1 the OTFT work involved realizing device models, using them to simulating circuits detecting features of EEG data. The OECT modeling effort focused on TCAD simulations and designing circuits of Muller C-elements based on experimentation and intuition. Finally, an important effort went into developing a new type of Bayesian inference that can be implemented using stochastic computing.
OTFTs: After extensive measurements of the pOTFTs in the low voltage regime, an important hysteresis was observed. The physical models in this regime were found to be not sufficiently developed. To surmount this, two empirical models were developed to enable circuit simulations. By the end of RP1, three models (physical and empirical) were functioning and had been used to design low transistor count circuits for BAYFLEX technology. The first testing of these circuits on a classification task has been done (see objective 3). Circuit layout for the OTFTs is well underway.
OECTs: For RP1, device characteristics of the OECTs were measured and TCAD modeling has been used to understand and fit this data. Physical compact models are not yet ready and first efforts at circuit simulations will focus on reproducing the characteristics using the models currently available in the literature. Discussions concerning the realization of Muller C-elements in OECTs led to first circuit implementations.
Bayesian inference using organic circuits: During RP1, a new method was developed for realizing Bayesian inference and stochastic computing.
Objective 2
The work on the fabrication part of BAYFLEX includes the realization of sensors for the measurement of EEG, of OTFT devices and circuits for the realization of stochastic bitstreams, and of OECTs for the classification task. RP1 focuses on demonstrating these three components separately and has proceeded as planned. During RP1, all technology partners have confirmed fabrication on an identical substrate in preparation for the partial and full integration planned during the project.
Concerning the cutaneous flexible sensors, the materials formulation and sensor design development have been established. In addition, standard EEG sensors from the EEG specialist were received by the partner developing organic flexible sensors were received and first tests to compare signal quality improvement, artifact content and durability are in progress.
Concerning the OTFTs, extensive characterization was done to provide sufficient data for the modeling in Objective 1. This involved a series of measurements containing around 30,000 data points for each of 200 devices of different sizes. In addition, a GEN2 technology is being developed for low power devices and Vt of ~-1 V (compared with -3—5V for Gen1) have been demonstrated
Concerning the OECTs, a process of record (PoR) was established for fabricating p-type, n-type, and ambipolar organic electrochemical transistors (OECTs). The goal was to ensure consistency, scalability, and repeatability throughout the fabrication process. Detailed characterization was provided for modeling and circuit simulation. We have established methods to efficiently tune the threshold voltage including: dual-gate structure, modification of composition of solid-state electrolyte (using KCl as an additive), the choice of the gate electrode material, and UV exposure. To work towards OECT classification circuits, we have realized cOECT inverters and demonstrated the high operational speed of operation.
Objective 3
The purpose of this objective is to demonstrate and benchmark the organic technologies for an electrophysiological task. The objective is divided into four parts: 1) classification, 2) demonstrator, 3) Life cycle analysis (LCA) and 4) benchmarking. The development of the demonstrator and its benchmarking begin RP2. In this reporting period, we focused on deciding on the most suitable classification task for the demonstrator, performing the life cycle assessment on OTFT and OECT devices (WP9) and showing how organic electronic devices can be used for simple classification tasks via circuit simulations based on the models of the organic devices. We have identified an appropriate classification task as well as a possible alternative.
The work on the model and simulate objective uses device measurements to develop models and circuits. During RP1 the OTFT work involved realizing device models, using them to simulating circuits detecting features of EEG data. The OECT modeling effort focused on TCAD simulations and designing circuits of Muller C-elements based on experimentation and intuition. Finally, an important effort went into developing a new type of Bayesian inference that can be implemented using stochastic computing.
OTFTs: After extensive measurements of the pOTFTs in the low voltage regime, an important hysteresis was observed. The physical models in this regime were found to be not sufficiently developed. To surmount this, two empirical models were developed to enable circuit simulations. By the end of RP1, three models (physical and empirical) were functioning and had been used to design low transistor count circuits for BAYFLEX technology. The first testing of these circuits on a classification task has been done (see objective 3). Circuit layout for the OTFTs is well underway.
OECTs: For RP1, device characteristics of the OECTs were measured and TCAD modeling has been used to understand and fit this data. Physical compact models are not yet ready and first efforts at circuit simulations will focus on reproducing the characteristics using the models currently available in the literature. Discussions concerning the realization of Muller C-elements in OECTs led to first circuit implementations.
Bayesian inference using organic circuits: During RP1, a new method was developed for realizing Bayesian inference and stochastic computing.
Objective 2
The work on the fabrication part of BAYFLEX includes the realization of sensors for the measurement of EEG, of OTFT devices and circuits for the realization of stochastic bitstreams, and of OECTs for the classification task. RP1 focuses on demonstrating these three components separately and has proceeded as planned. During RP1, all technology partners have confirmed fabrication on an identical substrate in preparation for the partial and full integration planned during the project.
Concerning the cutaneous flexible sensors, the materials formulation and sensor design development have been established. In addition, standard EEG sensors from the EEG specialist were received by the partner developing organic flexible sensors were received and first tests to compare signal quality improvement, artifact content and durability are in progress.
Concerning the OTFTs, extensive characterization was done to provide sufficient data for the modeling in Objective 1. This involved a series of measurements containing around 30,000 data points for each of 200 devices of different sizes. In addition, a GEN2 technology is being developed for low power devices and Vt of ~-1 V (compared with -3—5V for Gen1) have been demonstrated
Concerning the OECTs, a process of record (PoR) was established for fabricating p-type, n-type, and ambipolar organic electrochemical transistors (OECTs). The goal was to ensure consistency, scalability, and repeatability throughout the fabrication process. Detailed characterization was provided for modeling and circuit simulation. We have established methods to efficiently tune the threshold voltage including: dual-gate structure, modification of composition of solid-state electrolyte (using KCl as an additive), the choice of the gate electrode material, and UV exposure. To work towards OECT classification circuits, we have realized cOECT inverters and demonstrated the high operational speed of operation.
Objective 3
The purpose of this objective is to demonstrate and benchmark the organic technologies for an electrophysiological task. The objective is divided into four parts: 1) classification, 2) demonstrator, 3) Life cycle analysis (LCA) and 4) benchmarking. The development of the demonstrator and its benchmarking begin RP2. In this reporting period, we focused on deciding on the most suitable classification task for the demonstrator, performing the life cycle assessment on OTFT and OECT devices (WP9) and showing how organic electronic devices can be used for simple classification tasks via circuit simulations based on the models of the organic devices. We have identified an appropriate classification task as well as a possible alternative.
All objectives have obtained results beyond the state of the art during RP1.
Objective 1
1) the OTFT model in the low voltage regime was improved, 2) an empirical model of OTFTs using a multilayer perceptron regression was demonstrated, 3) TCAD modeling of OECTs was demonstrated, 4) circuits for BAYFLEX technology have been simulated based on the OTFT models of 1) and 2), 5) a Muller c-element using an OECT in an inverter simulation was proposed for the implementation of stochastic computing in organic technology.
Objective 2
1) Low voltage OTFT technologies have been developed, 2) Muller c-element in pOECTs have been demonstrated.
Objective 3
1) We have demonstrated by simulation in cadence based on physical models of organic transistors the implementation of a classification task involving EEG data using organic technologies. 2) We have realized a life cycle analysis at the level of single devices for OTFT and OECT technologies including new entries into the econinvent database for materials, flows and units.
Objective 1
1) the OTFT model in the low voltage regime was improved, 2) an empirical model of OTFTs using a multilayer perceptron regression was demonstrated, 3) TCAD modeling of OECTs was demonstrated, 4) circuits for BAYFLEX technology have been simulated based on the OTFT models of 1) and 2), 5) a Muller c-element using an OECT in an inverter simulation was proposed for the implementation of stochastic computing in organic technology.
Objective 2
1) Low voltage OTFT technologies have been developed, 2) Muller c-element in pOECTs have been demonstrated.
Objective 3
1) We have demonstrated by simulation in cadence based on physical models of organic transistors the implementation of a classification task involving EEG data using organic technologies. 2) We have realized a life cycle analysis at the level of single devices for OTFT and OECT technologies including new entries into the econinvent database for materials, flows and units.