cda-tum/sidb-defect-aware-physical-design

Design, layout, and simulation files of the paper "Atomic Defect-Aware Physical Design of Silicon Dangling Bond Logic on the H-Si(100)-2×1 Surface" by M. Walter, J. Croshaw, S. S. H. Ng, K. Walus, R. Wolkow, and R. Wille in DATE 2024.

Atomic Defect-Aware Physical Design of Silicon Dangling Bond Logic on the H-Si(100)-2×1 Surface

This repository provides supplementary data for the paper Atomic Defect-Aware Physical Design of Silicon Dangling Bond Logic on the H-Si(100)-2×1 Surface by M. Walter, J. Croshaw, S. S. H. Ng, K. Walus, R. Wolkow, and R. Wille in DATE 2024.

H-Si(100)-2×1 Surfaces

Of the hydrogen-passivated silicon surfaces used in this evaluation, two were experimentally fabricated in a lab and measured with an STM, the others were simulated based on experimental findings.

STM surface scans

The fabricated H-Si(100)-2×1 surfaces span a total of 830 × 652 and 740 × 1090 hydrogen sites, respectively, of which 8.57% and 6.26% are defective.

The respective STM surface scan files can be found in the stm_scans folder. The .xml files can be parsed by both SiQAD and fiction.

Fabrication process

The STM measurements were performed using an Omicron LT-STM system operating at 4.5K and ultra-high vacuum (3E−11 Torr). The STM tips were electrochemically etched from tungsten wire and sharpened using a field ion microscope. The used samples are highly arsenic-doped (≈1.5E19 atoms per cm³). They were prepared in-situ via resistive heating. To this end, they were first degassed at 600°C overnight followed by multiple flash annealing cycles at 1250°C. Finally, the samples were hydrogen-terminated at 330°C while exposing their surface to molecular hydrogen (1E6 Torr). The H₂ gas was converted to atomic hydrogen using a tungsten filament held at 1600°C.

The image acquisition was done using a Nanonis SPM controller with respective software. All images were taken in constant height mode with an imaging bias of 1.3V and a current setpoint of 50pA.

Simulated surfaces

We generated simulated surfaces of comparable size with variable defect rates of 1%, 0.5%, and 0.1%, including charged defects; and 5%, 1%, and 0.5%, with only neutral defects.

The surfaces have been randomly generated with the generate_defective_surface.py Python script.

The respectively resulting surface files can be found in the simulated_surfaces folder. The .txt files represent Python arrays and can be parsed by fiction.

Experimental Evaluation: Atomic Defect-Aware Physical Design

The experiments folder contains all data obtained by the physical design process laid out in the paper as well as a C++ code file that implements the algorithm to reproduce said data via the FCN framework fiction.

Source code: defect_aware_physical_design.cpp

The C++ code that implements the physical design algorithm presented in the paper. It utilizes the FCN framework fiction. To compile it, place the file in fiction's experiments folder and call CMake with the -DFICTION_EXPERIMENTS=ON flag.

To learn more, see fiction's documentation on how to build experiments.

NOTE: It might be necessary to adjust the file paths after copying the files into fiction's experiment folder.

Formatted data: results/

The folder contains raw data formatted as ASCII tables for all conducted experiments.

Considered logic networks for physical design

The respective logic networks that were used as specification for the defect-aware physical design process were taken from

A Placement and Routing Algorithm for Quantum-dot Cellular Automata by A. Trindade et al. in SBCCI 2016 (IEEE Xplore)

and

Placement and Routing by Overlapping and Merging QCA Gates by G. Fontes et al. in ISCAS 2018 (IEEE Xplore).

These networks are established benchmarks in the domain of FCN technologies and are available as Verilog files in fiction's experiment sandbox.