Nanoelectronics with the plenty of emerging devices and processes has been predicted to match in the years to come the relevance of Microelectronics (e.g. standard processes even at the nanoscale). However, physicists, chemists, technologists and engineers work on the demonstration of the feasibility of a single structure or device, oriented to their own discipline and using methodologies and models derived from their own background. There is no uniformity of approach nor a common methodology, every attempt remains
an attempt, and there is not a common vision about the possible direction to be followed. There is rarely attention to the usability of a real new device in a complex design, or, if an application is analyzed, very often results are not realistic, because methods and tools available for standard processes can only partially be inherited for emerging devices and computing principles. Basically, there is not a nanoelectronics common denominator.
Microelectronics experts that look at the future and make attempts to face nanoelectronics are slowed down, if not immobilized, by the lack of knowledges at the nanoscale that must be borrowed by other disciplines and, especially, they are defeated by the fragmentation of solutions, methodologies, simulators and types of experiments. And the interesting element is that several issues that arise when dealing with different possible technologies are pretty similar. This suggests that a comprehensive approach might give a boost to the nanotechnology era.
We have addressed these problems approaching the study of a few emerging devices and related circuits. This contribution focuses as a case study on Field Coupled Nanocomputing, and more specifically on the molecular version (MFCN).
As the application is Nanocomputing the goal is to verify whether it really enables computation on a large scale. And we approach this point with successive approximations.
A) As a first step we conceive a rough model of the device, take into account the most important technological constraints and also include design and process parameters with an approximate style. Exploiting the modularity and flexibility of Hardware Description Languages (that also include multiphysics) we implement a few complex architectures (examples: algorithms for biosequence analysis, methods for error correction codes in telecommunications, systolic arrays for arithmetic calculations, architectures for breast cancer detection, cryptographic application). During both the design and the verification phases methods are applied to solve issues arising from technological constraints (e.g. FCN is intrinsically pipelined and this creates several problems) or to better exploit the technology (e.g. systolic arrays, interleaving, optimal arithmetic solutions not possible with standard technologies ...). Results and applied methods allow to find out the general potential of the technology, or the total unfeasibility due to the explosion of a problem that becomes evident when a huge number of devices and design constraints (even rough) are taken into account (e.g. side effects in terms of area, power consumption, very high latency...). If the general results of this phase is positive then it is worth proceeding with the study.
B) If so, then a more realistic design should be considered. For this we developed a new CAD tool, ToPoliNano, which has two main features. The first is to enable a real Place & Route of a circuit, either using a full custom approach or an automatic one starting from an HDL description of the architecture. In both cases technological constraints must be accurately followed, and this means that a study of how devices are placed, interfaced, connected, powered, clocked... must be performed and validated, either using physical level simulators or experimental results. Here we show some examples of automatically placed and routed circuits based on FCN according to some of the possible implementations. C) The second main feature of ToPoliNano is the verification of the placed cells in the circuit structure. This means first of all to study a model of the device and of the basic cells or logic blocks. Several level of details can be accepted and implemented, depending on the complexity of the circuit an on the level of accuracy that is needed. These models should start and should be verified either with physical simulators when available or with experiments. Here we focus on the analysis of a single molecule for MFCN, on its transcaracteristic, pointing out how the main issue is the need to literally conceive a new way to study and model the device and to verify it.
D) The verification requires not only the model of a single device, but also the model of the interaction among devices, that is not so straightforward as it might seem, as very often at the molecular scale, and specifically for MFCN, a totally new algorithm for information propagation modeling has been conceived, because nothing existed and beacus totally different w.r.t. any other device based on conduction. Moreover the verification should include faults, process variation, noise, and even technological dependent parameters. So Models and data from technology and from experiments are to be developed and included in order to allow a more realistic verification of the placed and routed circuit. Here we show some results on simple circuits considering information propagation, presence of process variations and noise for the molecular implementation we are focusing on. We also show preliminary results on on going experiments for the demonstration and verification of the MFCN functionality.