In conventional chemical synthesis or chemosynthesis, reactive molecules encounter one another through random thermal motion in a liquid or vapor. In a hypothesized process of mechanosynthesis, reactive molecules would be attached to molecular mechanical systems, and their encounters would result from mechanical motions bringing them together in planned sequences, positions, and orientations. It is envisioned that mechanosynthesis would avoid unwanted reactions by keeping potential reactants apart, and would strongly favor desired reactions by holding reactants together in optimal orientations for many molecular vibration cycles. Mechanosynthetic systems would be designed to resemble some biological mechanisms.

While the description of mechanosynthesis given above has not yet been achieved, primitive mechanochemistry has been performed at cryogenic temperatures using scanning tunneling microscopes). So far, such devices provide the closest approach to fabrication tools for molecular engineering. Broader exploitation of mechanosynthesis awaits more advanced technology for constructing molecular machine systems - including a molecular assembler or precursors thereof.

Much of the excitement regarding mechanochemistry regards its potential use in automated assembly of molecular-scale devices. Such techniques appear to have many applications in medicine, aviation, resource extraction, manufacturing and warfare. Most theoretical explorations of such machines have focused on using Carbon, because of the many strong bonds it can form, the many types of chemistry these bonds permit, and utility of these bonds in medical and mechanical applications. Carbon forms diamond, for example, which if cheaply available, would be an excellent material for many machines. It has been suggested, notably by K. Eric Drexler, that mechanosynthesis will be fundamental to molecular manufacturing based on nanofactories capable of building macroscopic objects with atomic precision. The potential for these has been disputed, notably by Nobel Laureate Richard Smalley, leading to a famous dispute between the two of them - see nanotechnology. The Nanofactory Collaboration, founded by Robert Freitas and Ralph Merkle in 2000, is a focused ongoing effort involving 23 researchers from 10 organizations and 4 countries that is developing a practical research agenda specifically aimed at positionally-controlled diamond mechanosynthesis and diamondoid nanofactory development.

In practice, getting exactly one molecule to a known place on the microscope's tip is possible, but has proven difficult to automate. Since practical products require at least several hundred million atoms, this technique has not yet proven practical in forming a real product.

The goal of mechanoassembly research at this point focuses on overcoming these problems by calibration, and selection of appropriate synthesis reactions. The first product to be built by these means will probably be a specialized, very small (roughly 1,000 nanometers on a side) machine tool that can build copies of itself using mechanochemical means, under the control of an external computer. In the literature, such a tool is called an assembler or molecular assembler. Once assemblers exist, geometric growth (copies making copies) could reduce the cost of assemblers rapidly. Control by an external computer should then permit large groups of assemblers to construct large, useful projects to atomic precision. One such project would combine molecular-level conveyor belts with permanently-mounted assemblers to produce a factory.

In part to resolve this and related questions about the dangers of industrial accidents and runaway events equivalent to Chernobyl and Bhopal, and the more remote issue of ecophagy, grey goo and green goo (various potential disasters arising from runaway replicators, which could be built using mechanosynthesis) the UK Royal Society and UK Royal Academy of Engineering in 2003 commissioned a study to deal with these issues and larger social and ecological implications, led by mechanical engineering professor Ann Dowling. This was anticipated by some to take a strong position on these problems and potentials - and suggest any development path to a general theory of so-called mechanosynthesis. However, the Royal Society's nanotech report did not address molecular manufacturing at all, except to dismiss it along with grey goo.

Current technical proposals for nanofactories do not include self-replicating nanorobots, and recent ethical guidelines would prohibit development of unconstrained self-replication capabilities in nanomachines.

The technique of moving single atoms mechanically was proposed by Eric Drexler in his 1986 book The Engines of Creation.

In 1988, researchers at IBM's Zürich Research Institute successfully spelled the letters "IBM" in Xenon atoms on a cryogenic copper surface, grossly validating the approach. Since then, a number of research projects have undertaken to use similar techniques to store computer data in a compact fashion. More recently the technique has been used to explore novel physical chemistries, sometimes using lasers to excite the tips to particular energy states, or examine the quantum chemistry of particular chemical bonds.

In 2003, Oyabu et al reported the first instance of purely mechanical-based covalent bond-making and bond-breaking, i.e., the first experimental demonstration of true mechanosynthesis -- albeit with silicon atoms, not carbon.

In 2005, the first patent application on diamond mechanosynthesis was filed.

See also molecular nanotechnology, a more general explanation of the possible products, and discussion of other assembly techniques.