Radial addressing of nanowires

John E. Savage, Eric Rachlin, André DeHon, Charles M. Lieber, Yue WuACM J. Emerg. Technol. Comput. Syst 2006 · DOI

TL;DR

This paper introduces "radial encoding," a novel method for individually addressing nanowires (NWs) in crossbar memory architectures. By using layered shells around NW cores, this approach offers improved robustness against manufacturing misalignments compared to existing axial encoding methods, while achieving comparable memory densities.

What Problem Does This Paper Address?

As nanowires (NWs) and carbon nanotubes (NTs) shrink to nanometer scales, they hold immense promise for creating ultra-high-density memory and logic devices. A key challenge in realizing these "nanocrossbars" is the ability to individually address each NW. This requires a mechanism to select specific NWs from orthogonal sets to read or write data. Traditional methods often rely on axial encoding, where NWs are differentiated by doping patterns along their length. However, the fluidic assembly processes used to create these nanostructures lead to random displacements, causing misalignment between NWs and the control wires (mesoscale wires, MWs). This misalignment can result in ambiguous states, where a NW is neither fully on nor fully off, compromising device reliability. The paper addresses this critical issue by proposing a new addressing scheme that is inherently more resilient to such misalignments.

Methods

The core innovation is radial encoding, which leverages the layered structure of "core-shell" NWs. These NWs have a lightly doped core surrounded by multiple shells of different materials. By selectively etching these shells, different NWs can be exposed to control signals from mesoscale wires (MWs) at specific locations.

The paper details two main types of radial decoders:

  • Linear Decoder: This decoder uses a dedicated MW for each unique radial encoding. If a NW has n shells and there are a possible materials for each shell, this can lead to a large number of MWs if n and a are large. However, it offers a straightforward implementation.
  • Logarithmic Decoder: This approach uses a more efficient scheme, particularly for single-shell NWs, by employing binary encoding. It requires fewer MWs by grouping materials based on binary representations. The paper extends this to multi-shell NWs by dividing materials into sets based on shell parity, further reducing the number of MWs needed.

The authors also explore hybrid codes, combining axial encoding in the NW core with radial encoding in the shells. While these offer some advantages, they are generally found to be inferior to pure radial or axial codes in terms of effective pitch.

Datasets

Not discussed in the paper. The paper focuses on theoretical methods and analysis rather than experimental results on specific datasets.

Key Insights and Ideas

  • Radial Encoding's Robustness: The key insight is that radial encoding, by relying on shell etching and MW placement over exposed cores, is inherently less sensitive to random lengthwise displacements of NWs compared to axial encoding. This significantly reduces the likelihood of ambiguous states.
  • Shells as Differentiators: The layered shell structure provides a powerful and flexible way to differentiate NWs. By carefully selecting shell materials and etching processes, a large number of unique NW types can be created.
  • Trade-offs in Decoders: The paper highlights a fundamental trade-off between the number of MWs required and the complexity of the decoding scheme. Linear decoders are simpler but use more MWs, while logarithmic decoders are more complex but use fewer MWs.
  • Hybrid Codes' Limitations: While hybrid codes are explored, the paper concludes they are generally inferior due to increased pitch and similar misalignment penalties to axial codes.

Key Results

  • The proposed radial encoding scheme is shown to be competitive with axial encoding in terms of effective pitch and memory density, while offering superior robustness against misalignment.
  • The paper presents concrete examples of radial codes and decoders, demonstrating how a manageable number of shell materials and shells can generate a large number of distinct NW types.
  • Analysis suggests that while axial codes might have a slight edge in effective pitch under ideal conditions, the manufacturing complexity and susceptibility to ambiguity make radial codes a more practical and promising alternative.

Limitations and Future Work

  • Increased Diameter: A noted disadvantage of radial encoding is that adding shells increases the NW diameter, which can reduce the overall density compared to ideal axial codes.
  • Material Constraints: The effectiveness of radial encoding relies on the availability of differentially etchable shell materials. The paper lists some possibilities but acknowledges that expanding this palette would be beneficial.
  • Manufacturing Complexity: While more robust to misalignment, the precise etching of shells still presents manufacturing challenges.
  • Future Work: The authors suggest that further research into increasing the number of differentially etchable shell materials could enhance the capabilities of radial codes. They also point to the potential usefulness of a shell for separating axially coded NWs during fluidic assembly, even if not used for primary encoding.

Summary updated

BibTeX
@article{savage2006radial,
  author = {John E. Savage and Eric Rachlin and André DeHon and Charles M. Lieber and Yue Wu},
  title = {{Radial addressing of nanowires}},
  journal = {ACM J. Emerg. Technol. Comput. Syst.},
  year = {2006},
  doi = {10.1145/1148015.1148018},
}

Loading related content...

Figure