|Example Nanomaterials from Talapin et al at the University of Chicago (link).|
The overall goal when entering nanotechnologies into the market is low-cost, high-volume manufacturability, but at the same time, the materials' properties must be highly reproducible within a pre-specified limit, which Kelly states cannot happen below the 3 nm limit when trying to make arrays.
The top-down approach to manufacturing, which Kelly states is limited, uses external tools to cut and shape large materials to contain many smaller features. Its alternative, the bottom-up approach, involves piecing together small units, usually molecules, to construct whole materials -- much like a jigsaw puzzle -- however this process is too unpredictable for defect-free mass production of arrays.
As outlined by Kelly, standard manufacture operates under what is called 6-sigma yield - which essentially means there is a manufacturing defect rate of only 0.00034%. Those extremely tight tolerances are simply not met by the respective manufacturing processes of most nanotechnologies. However, there can be two ways to (re)consider these limitations.
The first is that - given the need for the technology - the scales of production will follow. Moore's Law is a well-known example. In short, due to continuous improvement in fabrication equipment and manufacturing technology - much of which was not around when Moore's "Law" was first proposed - the number of transistors possible on a single chip has doubled every two years from the 1970's to the present day. Based on this, it is reasonable to propose that additional improvements in manufacturing technology will help raise future nanotechnologies out of the "unmanufacturable" category.
The second way to consider these limitations is with reference to another complex system - the human body - which also possesses a great deal of complex functionality and manufacturing. For example, our own DNA replication and copying processes would not meet 6-sigma yield if not for clever DNA proof-reading and correcting mechanisms. In addition, some degree of DNA error is manageable because redundancy exists. Redundancy might not sound like an ideal situation for large-scale fabrication, but it is a useful tool to help maintain proper system functionality even if a small percentage of components are defective. And, thermodynamically speaking, if there will be an unavoidable amount of error, then redundancy may be the path of least resistance. Furthermore, if you accept defects will exist - if not from manufacturing, then due to in-situ operation - the design process changes from a defect-free nanotechnology to one which may perhaps even benefit from a certain concentration of them - much like the way doped semiconductors operate and interact. But of course, this may not be possible for some nano-materials and their applications, e.g. quantum dots requiring precise confinement energies.
Ultimately, as Kelly writes, some nanotechnologies are unfortunately bound to remain on a lab bench:
Many results in nanoscience can be shown to be intrinsically unmanufacturable in terms of ideas for applications in electronic or optoelectronic components, and so will remain as scientific curiosities.
Time will tell which are the most manufacturable and which are so in a sustainable manner.
 MJ Kelly. Intrinsic top-down manufacturability. Nanotechnology, Vol. 22 (2011), 245303.