COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\\M0BASL30\M1BASI30\M2BASB30\M3SUP\.
C00012 ENDMK
C⊗;
\\M0BASL30;\M1BASI30;\M2BASB30;\M3SUP;\.
\F0
MOLECULAR MEMORIES


\J	The ultimate in memory is to store information in single
molecules.  Then a mole of the substance used could store 6 x 10\F323\F0
units of information, where the unit might be a bit or might be
more.  My interest in the matter came from trying to set an upper
limit on information density and so trying to prove that single
molecules could not be used, and finally coming to the conclusion
that they could.  Whether such a memory could have practical
speed is still questionable but an interesting question.

	One approach to a molecular memory is to use a regular
solid, say a crystal of some sort, and have a state change in
each element of the crystal used to store information.  I can't
Addressing would involve locating the specific molecule and
reading or changing it without affecting its neighbors.
prove this to be impossible, but I haven't been able to think
of a way of doing it.

	I have thought of one bit-per-molecule memory, and I
want to describe enough of it to give the general idea.  However,
it is not in a state to be regarded as a rival for development
memory, so I would like it to be considered as an argument that
a long range research program aimed at large memories should
encourage and organize some brain-storming on molecule-per-bit
memories and should fund any plausible proof-of-concept projects
if they aren't too expensive.  Let me point out that any
device that can deal with single molecules in whole moles
will stretch the ideas of what is scientifically possible,
even if the device itself is not of immediate practical value.
This, in turn, would start people thinking about other ways
of dealing with single molecules in this or other fields.

	At this hour, I have time to deal only with reading,
but writing involves similar ideas.  Information is represented
in a chain molecule, and each molecule contains its address
as a string of 0s and 1s, and each bit is physically represented
by one of two radicals at the corresponding point in the chain.
In this simplest form, there is only one bit of information
in the molecule which is represented by one of two radicals
(not necessarily from the same two) at the end of the chain.
Between each two information radicals is a punctuation radical,
always the same but different from the information radicals.
(You might imagine that the molecules are DNA or proteins, but
I contemplate a method of reading and writing that doesn't depend
on using enzymes, i.e. a complex molecule that moves along the
chain.  [Hmm. A simple radical that was attached to each molecule
and moved along the chain might do what I want, but that is
a different, though possibly better idea])

	The memory consists of a large number of molecules - up
to some fraction of a mole - in a jar of water.  No attempt is
made to keep track of the location of a particular molecule.
There should be only one molecule with a given address.  At least,
if there are several, they must all have the same data radical
on the end.

	Addressing is carried out by sequencing the system among
three macroscopic physical states called \F1zero, one, and comma\F0.
Between each two address bits (\F1zero\F0 or \F1one\F0) is a
\F1comma\F0, so a typical sequence is "0,1,0,0,0,1,1" etc.
Imagine each physical state to consist of irradiating the
solution with a microwave frequency (from a maser) that will
stimulate a particular reaction, but there are other possibilities.

	Besides the information carrying molecules in the solution,
there is a \F1substrate\F0 consisting of molecules that can
attach themselves to the address radicals to ?orm a side chain.
The condition for attachment of a substrate molecule to a 0
address bit is that the physical state be \F1zero\F0 and there be a substrate
molecule attached to the punctuation radical on the left.  The
condition for attachment of a substrate molecule to a 1 address
bit is that the physical state be \F1one\F0 and there be a substrate
molecule attached to the punctuation radical on the left.
The condition for the attachment of a substrate molecule to
a punctuation radical is that the physical state be \F1comma\F0
and there be a substrate molecule attached to the information
radical to the left.  The effect of the addressing sequence
will be to attach a side chain of substrate molecules to the
address chains that match the physical states as they occur.
Only the word addressed will have a full chain attached to it.

	Once addressing is complete there are several ways of
proceeding.  One is to have two more substrate types of molecules,
one of which starts a chain reaction from the end of a side chain and
a 0 information radical and the other starts a chain reaction from
a 1 information radical.  The chain reactions are allowed to proceed
until a macroscopically detectable effect occurs - for example, a
molecular resonance.

One of two resonances occurs according to whether the information
bit is a 0 or 1.

	Finally, as a cleanup, the system is put into a physical state
that causes all the side chain bonds to break and the product of the
chain reaction to decompose so that we are back to the original state.

	Well, this complicated mechanism may seem unlikely to work
exactly as described, but the reader who inclines to the idea that
it is impossible to make a molecule-per-bit memory should try to
prove it impossible.  In the course of trying to debug his proof, he
may invent a better scheme than this one.

	[Returning to the bracketed afterthought, it may be easier
to make a radical move down the chain in accordance with the
addressing sequence of physical states than to build a side chain.]\.