## Deliverables

A novel flash memory has been developed by inserting self assembled InGaAs quantum dots into the spacer of a GaAs/AlGaAs modulation doped heterostructure which was processed by high resolution electron beam lithography and subsequent wet etching to define a narrow constriction along two side gates. The quantum dots were charged and discharged by applying positive and negative gate voltages to the side gates. Reading the device at zero gate voltage resulted in two different currents representing the charged and the discharged dot state. We have shown the memory to work with hysteresis of up to 1.7V for the single quantum dot layer and up to 4V for the double quantum dot layer. The single layer quantum dot flash memory was tested up to 270K and exhibited a write and erase time of smaller than 25ns. The double quantum dot flash memory worked even at room temperature and up to frequencies of 50kHz.
In contrast to state of the art devices due to direct tunnelling between quantum dots and the two dimensional electron gas the quantum dot flash memory is expected to have less degeneracy effects. Also the stacked quantum dot layer excels by an increased hold time.

The NANOTCAD package has been developed as a hierarchical set of tools. All tools have been prepared and freely delivered to the nanoelectronics community through the PHANTOMS Simulation Hub (http://www.phantomshub.com) or, in the case of one code using proprietary routines (SIMNAD), are freely available to those institutions with a licence of ISE-TCAD. Comprehensive manuals have been prepared for all tools and available for download on the project site (nanotcad.iet.unipi.it) and on the PHANTOMS hub. In addition, a tutorial session on tools developed within NANOTCAD will be organized during the NID Workshop in Cork, and a presentation with a tutorial part on NANOTCAD tools will be given at the International School MIGAS ``Towards Nanoelectronics'' held in Autrans 14-20 June 2003.
A series of collaborations have been established with experimental groups not participating to the present project, in order to validate the tools with a broader set of experimental data and to make the results available to the nanoelectronics community already before the end of project.

Three programs have been developed for the simulation of semiconductor nanostructures in quasi-equilibrium conditions in one-, two-, and three-dimensional domains (NANOTCAD1D, NANOTCAD2D, NANOTCAD3D, respectively). All codes are based on the solution of the many-body Schrodinger equation with density functional theory, local density approximation, and allow subdivide the domain in several regions with different types of quantum confinement, providing a reasonable level of flexibility. In addition, NANOTCAD2D also allows simulating ballistic FET both in the III-V and in the Si-SiO2 material system.
During the project duration such codes have allowed to simulate several nanoelectronic devices fabricated within and outside the project, and to gain important insights into the transport mechanisms of such devices.
Particularly interesting results have been obtained in the simulation of quantum dot flash memories, both in the silicon-silicon oxide and in the AlGaAs-InGaAs material systems, of single electron transistors, of silicon-germanium quantum wires and of ballistic field effect transistors.
With respect to initial project objectives, several additional results have been obtained. Indeed, the simulation of nanoscale field effect transistors was not present among original objectives, but was included in the course of the project since after the first year it was very clear that the formalism and the tools developed for other ballistic devices such as quantum point contacts could be easily extended to include FETs. In addition, a model for mesoscopic transport in the presence of decoherence, that was not an original objective of the project, was developed in order to obtain a better agreement with experiments as far as the simulation of transport in coherent or quasi-coherent devices is concerned. On the other hand, the three-dimensional codes are not as fast as initially expected. This aspect actually limits the possibility of using them as a design tool, since it makes a detailed exploration of the design space unpractical.
Sinergy with other european and national projects has been very fruitful for access to experimental data, in particular with the EU project ADAMANT, aimed at the development of silicon nanocrystal memories, and mainly focused on large scale fabrication and reliability aspects, and with the Italian Ministry of Research project "Single Electron Devices".
If nanotechnology will acquire industrial and economic relevance, it will strongly depend for its development on reliable Computer Aided Design tools, in the same way as Microelectronics relies upon TCAD tools. In that case, a broad basis of expertise in the development of CAD tools for nanotechnology - firmly established in Europe - would represent a real competitive advantage, with significant impact in terms of economic development and employment. NANOTCAD codes are tools for research and prototyping, but we believe their development has helped creating the necessary expertise on which a possible industrially oriented successor of the NANOTCAD project could be based.

On top of a high electron mobility transistor a narrow 25 nm gate has been deposited approximately 90nm away from the two dimensional electron gas beneath the surface. A universal signature for ballistic transport has been discovered which was characterized by a double peak structure of the derivative of the transconductance in respect to the gate voltage applied to the top gate electrode. Thereby ballistic transport has been reliably identified for the first time in macroscopic structures even at room temperature.

Two-terminal ultra-small devices comprising single phenylene-based molecules (benzene ring with two functional groups), which represent the smallest p-conjugated molecular device components, have been fabricated. The corresponding low temperature electrical transport characteristics revealed features that are tentatively ascribed to tunnelling through molecular states.
To fabricate these junctions, molecule were placed inside an electrode gap with a length below 1nm. Towards this end, two different electrical contacting approaches were further developed. In both cases, gold/palladium (AuPd) nanowires were employed as basic constituents. The AuPd wires were created via a template-based method using a transition metal oxide nanowire as removable etching mask to pattern the underlying AuPd layer. This allowed the fabrication of homogeneous wires of well-defined dimensions (~6nm in height, 15nm in width, and lengths up to 3µm) and with clean surface, which cannot be directly achieved by e-beam lithography techniques.
The first method consisted of incorporating the molecules between two nanowires arranged in crossed geometry, resulting in sandwich junctions of ultra-small area (~225nm{2}). To avoid the formation of short-circuits upon attachment of the top wire contact, an electrochemical method to assemble the molecules onto the bottom wire was introduced and optimized. Noteworthy, the electrodeposition conditions are such that the pi-conjugated part of the molecules is not affected.
Within the second approach, the molecules were contacted inside nano-gaps created within the AuPd nanowires by electromigration-induced breaking. Coating of the nanowires with the molecules prior to breaking proved useful to trap the molecules inside the nano-gaps. For that purpose, the same electrodeposition technique was found to be very effective in assembling the molecules onto the wires.
Although all investigated junctions showed distinct features in the electrical transport characteristics, problems were encountered due to a low reproducibility between different samples. This outcome is considered as a general, intrinsic problem in contacting single molecules that may not always have been sufficiently stressed in former works. A further complication was faced with respect to the possibility of controlling the current through the junctions. In effect, none of the studied samples revealed any gate effect on conductance, indicating the strong shielding of the molecules located between closely spaced electrodes.
The electrodeposition method offers significant potential for other, related applications, since it works with various types of mono- and bifunctional molecules. Moreover, it allows the formation of films consisting of an inert matrix that may include active molecules or clusters in an appropriate concentration. Thus, this method may be of interest for the fabrication of molecular devices, such as memory arrays based upon crossbar structures, which comprise molecules that are usually hard to assemble by standard methods (self-assembly, Langmuir-Blodgett films, spin coating).

IMNAD (SIMulator for NAnoDevices) is a quantum mechanical 3D simulator for semiconductor devices based on a temperature-dependent effective mass formulation of density functional theory. It can be used to compute the self-consistent quantum mechanical charge density in semiconductor nano-structures, and can handle both direct gap materials (like III-V semiconductors) and materials with a silicon-like six-valley band structure and anisotropic effective mass. SIMNAD can calculate the full 3D wave function in MOSFET channels and SET leads based on the scattering matrix method with open boundary conditions. This can be used to compute the tunnel probabilities in SETs. Conductances computed by SIMNAD are either tunnelling conductances (e.g. in single-electron transistors) or quantum-ballistic conductances (e.g. in quantum point contacts or ballistic MOSFETs). SIMNAD uses a conductance model for Coulomb blockade devices based on the description by Beenakker (sequential tunnelling). The program can also compute the coherent resonant tunnelling current based on a modified Landauer-Buttiker formula.
With a coupling-enabled version of the semi-classical device simulator DESSIS-ISE one can run simulations in coupled mode, i.e. SIMNAD will compute the quantum mechanical charge densities for band profiles provided by DESSIS-ISE and will communicate this data back to DESSIS-ISE for further processing (e.g. for use in a DESSIS-ISE self-consistency iteration). This enables the simultaneous modelling of a (3D) quantum-mechanical charge distribution in a sub-region of a larger device, which is under full operation. Such a methodology is unique and no other comparable simulation package is known. So far, only a non-self-consistent version is available, but the self-consistent version is being developed and tested. In the non-self-consistent version of the coupling scheme, the quantum-mechanical charge density from SIMNAD is frozen-in for a subsequent DESSIS-ISE solve without any feedback. In the self-consistent version, the Poisson solver of SIMNAD is deactivated and the quantum-mechanical charge density is computed with the fixed potential obtained in a DESSIS-ISE run with the previous quantum-mechanical charge. The iteration is continued until convergence. The two simulators communicate via semaphor files, where DESSIS-ISE requests an update of the quantum-mechanical charge if required.

The development of a software tool can be divided into two developments: the newly developed theoretical method for describing correlated electronic transport, and the software code which implements the method. The goal of the software development was to create a design tool for predicting current voltage characteristics for material systems on the sub 10 nanometer scale. This is to serve as a basis for setting standards for the computations on this length scale and to allow for development of accurate approximations for handling electronic device design on the nanoscale.
Initially the code will be offered for evaluation through the Phantoms software hub, and access is open to all potential users. Development of the code will continue with emphasis on ease of use, scalability and further integration into conventional technology computer aided design packages.
Commercial exploitation of the code requires the incorporation of the code into existing software packages, and this already includes the NANOTCAD package. If the method can be made sufficiently user friendly (in terms of integration into a GUI), the code can be used in commercial device design for end of the roadmap technologies and beyond. Initially the code will be useful for evaluating alternative technologies, but as device dimensions scale to the 20nm range and below, this programme offers the potential for accurate device design on an atomic scale.

S The Wigner-Boltzmann equation for electrons in semiconductor devices is solved numerically by means of a novel Monte Carlo method. The equation describes both quantum interference and dissipation effects due to carrier scattering. The methodology for deriving the method is summarized in the following:
The integral form of the Wigner-Boltzmann equation is used as a starting point for deriving the method. The kernel of the adjoint equation has been decomposed into a linear combination of conditional probability densities. These densities represent the transition density used for the construction of numerical trajectories. The properties of the transition density employed allow a particle picture to be introduced. In this picture dissipation and interference phenomena are taken into account by two alternative processes involving quasi-particles. Dissipation caused by interaction with phonons and other scattering sources is accounted for by drift and scattering processes corresponding to the semi-classical Boltzmann transport picture. Interference effects due to the Wigner potential are associated with generation of pairs of particles having statistical weight +1 and -1. The classical force term is separated from the Wigner potential and included in the Liouville operator. With this modification, the developed model corresponds to a Boltzmann equation augmented by a generation term. The challenge of employing such method is to handle the avalanche of numerical particles properly. The problem has been solved for stationary conditions: Particles of opposite weight and a sufficiently small distance in phase space are continuously removed in the course of a simulation. The cancellation is due to the fact that such particles have a common probabilistic future but opposite contribution to the statistics.
Experience about the properties of the MC method has been collected by simulation of different types of resonant tunnelling diodes. The novel MC method has been validated by comparison with NEMO-1D and comparison with measurements. In some cases phonons scattering is found to play an important role, as predicted by both simulators.
The novel MC method has been integrated in the one-dimensional device simulator VMC (Vienna Monte Carlo code). A description of the concepts, features, input deck and out files of VMC are described in a user's guide. The basic input quantity is the profile of the conduction band edge in the device. This profile can either be specified analytically in the input deck, or, for the purpose of a self-consistent simulation, be read from an input file, typically generated by a Schroedinger-Poisson solver. The simulator VMC
is released by the Institute for Microelectronics, see http://www.iue.tuwien.ac.at. Access to the simulator can be found also on the phantoms simulation hub.