Due to some poor editing, there are a number of errors in the discussion.
As far as we can tell, the data used for actual calculations is correct,
but some of the example calculations given and one of the equations could
be misleading.
- P. 222, column 2, under equation 30.
It should say:
P_{cw}=0.95^{2} x 0.9999^{2899}=0.68
As written it raises to the 28990 power. The actual calculation is
formulated in terms of junctions, and there is one junction every 10nm.
So, the 28.99 micron wire has 28990nm and 2899 junctions.
- P. 222, column 2, later in the same paragraph as the above item:
d=2x82+1=165
The text incorrectly states d=...=185.
The conclusion N__{cdata}=625 is still correct.
The error was in the simple addition.
- P. 226, Equation 46. The left hand side should be
N_{cyield}-N_{ccdata}, so that we have:
The calculation is correct in the computer code which
calculated the plots, so this is simply a presentation
error.
- P. 226, paragraph (example) following equation 46.
N_{cdata}=185. Reference [34] provides
a [264,199,17] code which would provide N_{cdata}=199.
This still illustrates the intended point that the Gilbert
Bound used is conservative compared to achieveable codes.
We apologies if these errors make the paper more difficult to read.
Acknowledgments: Thanks to Dmitri Strukov for helping identify
and resolve these issues.