21 February 2000
(The following brief note extends the discussion of a preliminary
draft paper  on molecular assembly that was privately circulated in November
1998.)The overall problem of molecular manufacturing is to deterministically
construct vast numbers of almost exactly similar congeries of atomically-specified
parts, e.g. molecular machines.
A non-SPM-based approach to molecular
Although it would reduce the problem of parts fidelity and orientation during
assembly, it will be very hard to build molecular structures atom at a time
because individual atoms are more difficult to mechanically grasp and release,
and have less-easily controlled covalent reactivities than molecules. Thus early
molecular manufacturing systems should probably build using molecular parts
rather than atomic parts. Molecular parts having 10-100 atoms placed in precise
relative positions are already accessible to current chemical synthetic
techniques, and can in theory be produced in mole quantities, so there would be
an abundance of selfsame parts with which to work.
But molecular manufacturing also requires that one be able to assemble large
numbers of parts deterministically to make molecular-featured useful (e.g.
marketable) product. It is quite possible to imagine that an SPM could be used
to manually assemble a simple nanoreplicator device. The replicator device could
be free-floating or wall-anchored in design. The initial one-off replicator
would then operate on nanoparts presented to it and build more replicators. When
enough replicators had been built, a signal would be sent to the fleet of
replicators that tells them to change their mode of operation, and the
replicators now begin assembling useful product from the sea of parts, rather
than more replicators.
The practical problem here is that the design for the replicator must be
known in advance and completely specified. The replicator must be sophisticated
enough to preclude all alternative assembly pathways except the one or more
pathways which are desired. The replicator also must be capable to at least two
(or more) distinct modes of operation – as the structure of the useful product
object differs more and more from the structure of the replicator, the
replicator will need increasing complexity of structure and sophistication of
design to properly execute its two (or more) distinct output modes. Also, at
least from the viewpoint of simple early systems, such a replicator will be very
functionally brittle. It will be able to assemble only a very limited range of
output products, and slight changes in the design of those output products may
necessitate a complete redesign of the replicator structure, with systemwide
alterations possibly needed.
A less-brittle approach to mass-production molecular assembly would be a
technique that would allow alteration of each parts-assembly step without
necessitating a redesign of the entire system; which would allow new assembly
pathways to be inserted without redesigning the entire system; which would
isolate each step of the parts assembly process, thus reducing or eliminating
unwanted assembled-nanopart congeries.
Let's assume that we have some nanometer-scale parts that we wish to assemble
into useful product nanomachines in a deterministic fashion. At least three
different near-term ways to get them to assemble are readily apparent:
A benefit of this technique is that assembly operations can
be quickly and accurately performed on mole quantities of parts, thus permitting
massive parallelicity. One practical concern is that a different mechanoenzyme
may be required for each parts pairing that we desire to assemble. However, with
proper system design the number of necessary enzymes can probably be
significantly reduced because the parts-assembly space will probably reduce to a
relatively small number of distinct classes of binding sites, mechanoenzyme
backbone lengths, and actuator modes.
- Self-assembly in solution phase. Each part is manufactured so as
to incorporate complementary surface features or handles on one face, and
noncomplementary or nonattractive surfaces on all other faces, with the result
that parts which meet in random orientations or positions will join only in
one specific position and orientation. A collection of parts will thus spontaneously
assemble into the desired congeries or finished product nanomachines. The
principle difficulty here is that at each step, all alternative assembly pathways
must be known, well-characterized, and excluded in advance; as the number
of assembly steps increases, the total number of such pathways may become
very large, even if the self-assembly steps are performed strictly sequentially
allowing isolation of each assembly cycle.
- SPM-based assembly in vacuum or solution phase. Each part is grabbed
by a separately maneuverable functionalized SPM tip, brought into contact
in the proper relative orientation under the view of an SEM or similar viewing
facility, and the parts then bond (a) covalently if chemically active handles
have been included on the parts’ surfaces, (b) electrostatically if
the parts contained charged handles, or (c) by simple van der Waals forces
if the parts contain complementary surface features. The SPM tips then must
detach from the assembled unit, releasing the assembled unit for further assembly
downstream. One major difficulty here is that even with a large number of
of parallelized tips working in unison, the output count is still going to
be relatively low. This is certainly not fatal to the concept. Many useful
nanomachines can be conceived for which only one (or some small number) will
suffice in a particular application -- for example, sensors (at the low-complexity
end of the spectrum) and replicators (at the high-complexity end).
- Mechanoenzymes [2-4] in solution phase. A mole quantity of parts
"A" and "B" are suspended in working fluid. Parts "A" and "B" have no natural
complementarity and will not spontaneously self-assemble. Some number of identical
"mAB" mechanoenzyme molecules are next introduced. The mechanoenzyme molecules
have a reversible binding facility for part "A" at one end, a similar facility
for part "B" at the other end, and an actuator mechanism in between. As soon
as the mechanoenzyme is introduced, parts "A" and "B" attach to their proper
places and orientations on each mechanoenzyme molecule. A chemical signal
is then introduced into the working fluid, which causes the mechanoenzyme
molecule to flex and/or rotate, bringing parts "A" and "B" into contact in
a specific position and orientation with sufficient force to overcome potential
barriers and induce the mechanical joining of the parts. The binding facilities
for "A" and "B" are then disabled, allowing the compound part "AB" to be released
(and later recovered, separated and purified by sieving or centrifugation
or other similar means), and the mechanoenzyme is reverse-actuated, returning
it to its original cocked position and ready for another cycle of activity.
Both the binding facility release event and the mechanoenzyme reverse-actuation
event may be triggered by additional chemical signal(s), or might possibly
be triggered automatically following the prior mechanical sequence. Construction
of complex nanomachines is accomplished by employing an ordered sequence of
two-part suspensions into which parts-specific mechanoenzymes are introduced,
catalyzing each assembly operation, one after the other. Energy must be injected
into the mechanoenzyme at some point in its cycle, in order to drive the cycle,
and this may also serve as one of the aforementioned chemical "signals".
Useful product can include addditional mechanoenzymes, which themselves may
be relatively simple in structure (unlike replicators). Since in principle only
one (or just a few) mechanoenzymes are needed to catalyze the assembly step, you
could initially manufacture the mechanoenzymes as one-offs using SPMs or some
functionally similar technique. Mechanoenzymes could even be entirely
Pursuit of the mechanoenzyme pathway to molecular assembly must address at
least the following issues:
First, a research program to develop mechanoenzymes might first establish a
database of useful nanoparts which might be desired to be assembled -- such as
variously-shaped blocks, ratchet-shaped and tooth-shaped parts, hinges, toroids
and insertable rods of various lengths and diameters, and so forth.
Second, a reliable reversible binding mechanism for each part must be
devised. As long as each part has physically distinguishable faces this should
be possible. Since only two types of parts will be present at any time, the
competitive binding analysis should be much simplified from a situation in which
many different parts are simultaneously present. The strength of the binding
must be greater than the maximum forces expected to be exerted upon the parts
during assembly, unless the assembly force vectors are designed to be additive
with the binding forces during the assembly operation, as would be prudent to
do, if possible.
Third, a means by which the binding of the part to its receptor facility may
be externally controlled in a reversible fashion must be devised. Molecular
pumps might provide a possible analog – configurational changes make the binding
site active or inactive.
Fourth, a specific mechanism for achieving the externally-controllable mechanoenzyme
actuator motion must be devised. Examples of this might include butterfly molecules,
Ned Seeman's DNA manipulator device , a free-floating dynamin  pinch molecule,
or even two kinesin motors bearing parts and approaching each other by walking
along two adjacent oppositely-oriented microtubule tracks. Can these sorts of
molecular motors exert sufficient force to accomplish the desired assembly operation?
The kinesin motor stall force is 5-7 pN; using two opposed kinesin motors, one
bearing each part and walking towards each other, would allow a total 10-14
pN to be applied to the workpiece. (Note also that the kinesin family includes
two types, a plus-walker and a minus-walker . In principle, rigid crossbeams
affixed to multiple motor molecules on multiple tracks could provide whatever
summed forces might be deemed necessary at the workpiece.)
Fifth, some means of restoring the actuator to its starting position, and of
supplying the appropriate energy to drive the actuator motion, must be devised.
In the case of chemomechanical kinesin-based or dynamin-based nanomotors, or
related biological approaches, this could involve ATP or GTP.
- Robert A. Freitas Jr., "A novel pathway to first-generation
molecular assemblers," preliminary draft paper, 18 November 1998.
- M.S. Mooseker, T.R. Coleman, "The
110-kD protein-calmodulin complex of the intestinal microvillus (brush border
myosin I) is a mechanoenzyme," J. Cell Biol. 108(June 1989):2395-2400;
A. Garcia et al, "Partial
deduced sequence of the 100-kD-calmodulin complex of the avian intestinal
microvillus shows that this mechanoenzyme is a member of the myosin I family,"
J. Cell Biol. 109(December 1989):2895-2903.
- G. Steinberg, "A
kinesin-like mechanoenzyme from the zygomycete Syncephalastrum racemosum shares
biochemical similarities with conventional kinesin from Neurospora crassa,"
Eur. J. Cell Biol.73(June 1997):124-131.
- M.A. McNiven, H. Cao, K.R. Pitts, Y. Yoon, "The
dynamin family of mechanoenzymes: pinching in new places," Trends Biochem.
Sci. 25(March 2000):115-120; D. Danino, J.E. Hinshaw, "Dynamin
family of mechanoenzymes," Curr. Opin. Cell Biol. 13(August 2001):454-460.
- C. Mao, W. Sun, Z. Shen, N.C. Seeman, "A
nanomechanical device based on the B-Z transition of DNA," Nature 397(14
- Ulrike Henningsen, Manfred Schliwa, "Reversal
in the direction of movement of a molecular motor," Nature 389(4 September
1997):93-96, 15-16 (comment).
Last updated on 2 November 2002