Crystal polymer has a characteristic phase change from crystal to
amorphous at temperature Tg. After the phase change, the mechanical
shape can be flexible. If a tip is pushed on the polymer after the
phase change, the polymer will leave a dip after it is cooled down.
The polymer is deposited on Si substrate. The resistance between the
substrate and the tip depends on the material between them. If there
it is a dip mark (written state), it is air (or other gas used) plus
remaining polymer as the medium, a read can be made by checking the
resistance of the medium. The written sate can be erased (recovered)
by heating surrounding area, which requires certain buffer space
between neighbor cells. This is the principle used in Millipede to
store information in a flat thin polymer layer.
IBM targeted a high density, rewritable, non-volatile Millipede
memory product (internally?) on 2005. The following is quick
comments about the technology.
1.It is just two dimensions (xy); Multi stack architecture (stack in
z direction) is impossible. This limits its density per square.
2.It requires certain buffer space for rewritable memory; this also
limits its density.
3.It may achieve multibits per cell by controlling the depth of the
dip, however, it is not easy, IBM did not mention that. This limits
4.Uniform thin polymer thickness is critical for high density.
However, control this uniformity in nm range is not easy for current
technology. This may give issue on reliability of large (area) memory.
5.Rewritable Millipede memory means the polymer does not wear out
after many (>10^6) burning cool cycles. It is not easy to find a
report from IBM how many cycles the polymer material can be operated.
Clearly, reliability for rewritable memory is a potential issue.
6.Median present between polymer and the tip is air in IBM report. As
we know most of polymer can be oxidized in air. If the other inert
gas is used, package will increase the cost.
7.The multi tip is used in order to have fast speed. As we know very
fine AFM tip plus the control circuit is not cheap. The cost of the
product cannot be low in short term.
Sure, we may find the other good and bad aspects. Overall, the risk
level of the technology is ranked at Four Stars out of Five