Video: A digitally controlled
molecular machine
(>3 billion years of patchwork)

A few weeks ago, I highlighted some of the most popular posts in Metamodern’s first year (see “Knowledge about Knowledge…”). Posts that offer videos, documents, or talk slides also ranked high:

With downloadable documents and talk slides:

• Molecular Nanomachines: Physical Principles and Implementation Strategies
• My MIT dissertation — a draft of Nanosystems — is now online
• Slides for Talk on Nanotechnology and Computational Challenges
• Slides for Berkeley Talk on Molecular Nanosystems
• The Technology Roadmap Translated: Russian

With molecular videos:

• Productive Nanosystems: The Movies
• Productive Nanosystems: The Ribosome Videos

With videos of machines making things at blazing speeds:

• High-Throughput Nanomanufacturing: Small Parts
• High-Throughput Nanomanufacturing: Assembly
• High-Throughput Nanomanufacturing: Assembling larger products

• Knowledge about Knowledge:
  The most popular posts in the first year

A challenge question:
Measured by cumulative traffic, one post on Metamodern ranks well above the rest. It was neither very recent, nor very early, nor on any of the topics above. Which one is it, and why is it popular? (Actually, I’m baffled by the second part of the question.)

“Big Dog”: An agile quadruped robot
Climbing a hill.     Slipping on ice.
Climbing down.     The leap.
(See potentially disturbing video)

---

Brought to you by Boston Dynamics and DARPA

A fast, dexterous robotic hand
The hand.    Motion blur.
The toss.    The catch.
(See potentially disturbing video)

---

Brought to you by the Ishikawa Komuro Laboratory
at the University of Tokyo

Forget about clumsy, lumbering robots.
Think fast, precise, and acrobatic.

[Update: only potentially disturbing videos.]

• — Small Parts (with videos, no robots)
• — Assembly (with videos, no robots)
• — Assembling larger products (with video of a very fast robot)

And…

• Why I hate “nanobots”

The ribosome diagrammed

---

The ribosome in reality
(see video below)

While browsing the literature on the catalysis of bond formation in protein synthesis by ribosomes*, I came across a wonderful set of videos of the ribosomal protein manufacturing system at work, shown in recent-state-of-the-art molecular detail. These videos were presented in a Chemical & Engineering News article online, but I missed seeing them at the time. (This shows a shortcoming of my habit of reading only the paper bundles that the ACS sends by post.)

…but beware of
Ornithological Aeronautics

Ribosomes are productive nanosystems, that is, machines that combine small molecular building blocks under digital control to build complex structures. They’re programmable by means of genetic engineering, and extraordinarily useful in emerging molecular technologies. However, when watching how ribosomes work — in strange, messy, and subtle ways — it’s wise to keep in mind the usual relationship between biological examples (horses, birds) and engineered, mechanical systems (cars, aircraft) that serve similar functions. Biological examples show that functions can be performed (ground transport, air transport), but they don’t necessarily suggest the best approach to building machines for human purposes. If aeronautical engineering had been a branch of ornithology, there would be extensive research on artificial feathers and the ongoing challenge of artificial muscle, but no passenger aircraft.

There’s a much learn from ribosomes, but also much that needn’t be imitated.

Video: A digitally controlled
molecular machine
(>3 billion years of patchwork)

The first video (from the MRC Laboratory of Molecular Biology in Cambridge, UK) is based on structures determined by x-ray crystallography and electron microscopy; these are animated, and rendered at a fine, molecular level of detail. The video shows the entire process of protein manufacture, from initiation to release. The frantic pace of the middle part of the video shows action in approximately real time.

Note, however, that the actual motions of these biological molecular machines are far less purposeful — depicting realistic molecular trajectories would either show a blur, or be excruciatingly boring. (See “The videos lie because they must” in Molecular Machine Assembly: The Movie.)

Video: A smoother view
of the elongation cycle

The second video, produced by researchers at the Howard Hughes Medical Institute, focuses on the repeating cycle that extends the amino acid chain. It, too, is based on real structural data, but is presented at more diagrammatic level of detail.

Video: Molecular dynamics
of tRNA binding
(>2 billion years; almost no change)

The final video, showing detail at a molecular level, is shows the results of a targeted molecular dynamics simulation; targeting is another way to give an unnatural degree of direction to molecular motion, avoiding both excruciating boredom and (here) consumption of inordinate amounts of computer time.

The video shows a key step in tRNA binding, the “accommodation” process, the step where the appendage carrying the next amino acid (green) swings across from right to left. This step is central to kinetic proofreading in ribosomal translation, a process that keeps error rates tolerably low. A landmark paper on this process was co-authored by the new U.S. Secretary of Energy, physicist and Nobel laureate, Steven Chu.

• Productive Nanosystems: The Movies
• Molecular Machine Assembly: The Movie
• Nanomachines: How the Videos Lie to Scientists

And of high-throughput manufacturing:

• High-Throughput Nanomanufacturing: Small Parts (with videos)
• High-Throughput Nanomanufacturing: Assembly (with videos)

A simple molecular chain

---

A 2D simplification of
its typical complex,
disordered, 3D structure

A scientist recently remarked to me that molecular modeling techniques cannot accurately predict the mechanical properties of typical polymers, even one as simple as polyethylene, a hydrocarbon consisting of long chains of –(CH₂)– units. He was, I think, suggesting that molecular modeling may tell us little about molecular technologies based on structures that would be far more complex.

It’s important to understand why this plausible line of reasoning is mistaken.

Where can molecular mechanics models give accurate predictions? One area includes precisely structured solids with stable, covalent bonding and no conformational degrees of freedom. Ideally structured, hydrogen-terminated silicon crystals provide examples, as does a large class of existing and potential polycyclic structures with modest strain and saturated bonds. The elements H, C, N, O, F, Si, P, S, and Cl provide a broad palette, and the higher adamantanes are hydrocarbons in this class.

[ Note: When operation of a device involves only low stresses and strains, the elastic responses of structures of this sort typically are nearly linear, which greatly simplifies (for example) calculations of entropic contributions to free energy.]

Where do molecular mechanics models encounter difficulties? One area includes disordered solids with extensive non-covalent interactions and conformational degrees of freedom, and polyethylene is a good example. Its deceptive simplicity stems from the simplicity of its molecules; its challenging complexity stems from the organization and interaction of those molecules.

At room temperature, polyethylene forms a disordered structure consisting of of small crystallites threaded by multiple, partly-folded chains. Under increasing tension, chains unfold and slide, distributing tension unevenly and breaking in more-or-less random patterns. The mechanical properties of the material (for example, stress-strain curves and maximum elongation to failure) depend on polymer chain lengths and processing history: both milk jugs and plastic bags are commonly made of polyethylene, but so is Dyneema, a polyethylene material in which the same repeating units — but in longer, highly oriented chains — form fibers that rival high-strength steel.

In short, polyethylene forms complex, disordered materials that are quite unlike well-ordered proteins or other components suitable for use in atomically precise systems. Noncrystalline solids that form spontaneously from simple molecules will often be more difficult to model than even quite complex molecular machinery, in part because they have no specific structure.

This example illustrates aspects of the contrasting perspectives of scientific inquiry and engineering design. I’ve recently written about the methodology of exploratory engineering as a basis for applying the predictive power of scientific knowledge to achieve a limited — yet powerful —set of insights into future technologies, and in an earlier post I described why the intricate molecular systems that can be made today are challenging to model, and why the intricate molecular systems that are easiest to model cannot (yet) be made.

I see many steps between where we are and what we can see ahead, several turns further along a winding road. I expect bionanotechnology to play a central role, and for design, biomolecular modeling is already surprisingly capable.

• Making vs. Modeling:
  A paradox of progress in nanotechnology
• Exploratory Engineering:
  Applying the predictive power of science
  to future technologies
• Macromolecular Modeling for Molecular Systems Engineering
• The Physical Basis of Atomically Precise Manufacturing

Tool maker


Innovative tool maker

Some scientists have expressed surprise that Homo floresiensis made and used stone tools despite having remarkably small brain. I can see two reasons why this should be no cause for astonishment: One is the intelligence of crows, the other is the Baldwin Effect.

Crows

An adult H. sapiens brain typically weights well over a kilogram. H. floresiensis made do with about 1/3 as much brain mass, while Betty (Corvus moneduloides) has only 1/100th of ours.

H. floresiensis and its ancestors made and used stone tools, apparently without change, for many times the span of recorded history. Betty, a New Caledonian crow, not only makes and uses tools, but found a way to make them from a new material.

New Caledonian crows make hooks from tree branches. On her 5th encounter with various pieces of wire and a puzzle requiring a hook, Betty bent a straight wire to make a hook. Her Oxford research group has many videos, including this video of Betty’s 7th trial. The tasty tidbit is in the bucket at the bottom of the transparent tube. She later found better ways to bend the wire.

If a crow can discover how to make tools in new ways, I find it hard to be surprised that a hominin with 30 times more brain mass could learn to make tools in old ways, even though making the tools required more steps.

The Baldwin Effect

Also, evolution can be influenced by learning. If some members of a species can find a solution to a survival-related problem (whether by insight, imitation, or chance), then there is apt to be an advantage if they can learn the solution more easily. The resulting evolutionary pressure is called the Baldwin Effect, and it can compile the learned behaviors of one generation into the inclinations or even instincts of a later generation.

The Baldwin Effect would have helped the ancestors of H. floresiensis to continue imitating their parents, even with moderately smaller brains. It’s an interesting phenomenon, and it deserves mention.

…and spiders, too

If we made complex nets like those of orb-weaving spiders, we’d very likely imagine that this required brains with hundreds of grams of smart neural tissue. But can the brain of a small orb weaver be as much as a milligram — 1/100,000 of ours?

Molecular manufacturing
(some development required)

---

Montage of frames from the Productive Nanosystems video by John Burch

While upgrading parts of the E-drexler.com website, though, I’ve been re-reading some of the on-line content from Nanosystems: Molecular Machinery, Manufacturing, and Computation, the book that grew into, then out of, my MIT dissertation. Nanosystems explores what physics tells us about the potential of advanced molecular manufacturing systems and products. It outlines some ideas about pathways (the main focus of my blog posts), but the chief aim of Nanosystems is to examine methodologies for surveying technologies beyond our reach, and to describe some of what can be seen in the broad territory that will be opened by progressive advances in atomically precise fabrication.

The book begins like this:

Chapter 1

Introduction and Overview

1.1. Why molecular manufacturing?

The following devices and capabilities appear to be both physically possible and practically realizable:

• Programmable positioning of reactive molecules with ~0.1 nm precision
• Mechanosynthesis at >10⁶ operations/device ? second
• Mechanosynthetic assembly of 1 kg objects in <10⁴ s
• Nanomechanical systems operating at ~10⁹ Hz
• Logic gates that occupy ~10^–26 m³ (~10^{– 8} μ³)
• Logic gates that switch in ~0.1 ns and dissipate <10^{– 21} J
• Computers that perform 10¹⁶ instructions per second per watt
• Cooling of cubic-centimeter, ~10⁵ W systems at 300 K
• Compact 10¹⁵ MIPS parallel computing systems
• Mechanochemical power conversion at >10⁹ W/m ³
• Electromechanical power conversion at >10¹⁵ W/m ³
• Macroscopic components with tensile strengths >5×10¹⁰ Pa
• Production systems that can double capital stocks in <10⁴ s

Of these capabilities, several are qualitatively novel and others improve on present engineering practice by one or more orders of magnitude. Each is an aspect or a consequence of molecular manufacturing.

The table of contents and sample chapters can be found here. Together, they give a damn good overview of the topic and how it fits with the rest of science and engineering. There’s a glossary, too.

A U.S. National Academies study reviewed the scientific basis of molecular manufacturing, primarily referencing Nanosystems; the committee recommended funding experimental research. Links can be found here.

Replacing tin atoms
with silicon
using an AFM

“Complex Patterning by Vertical Interchange Atom Manipulation Using Atomic Force Microscopy”
Y Sugimoto, et al., Science, 322: 413–417 (2008).

Shortly before I launched Metamodern, Science published a remarkable paper by Sugimoto et al. describing atom-by-atom manipulation of a monatomic layer of tin (Sn) on silicon (Si). The animation to the right shows the steps in constructing a pattern of Si atoms that spells ‘Si’. Each frame is an atomic force microscope image made with the same tip used to construct the pattern, part of a commercially available silicon cantilever.

The process works by interchanging atoms between the tip and the surface. Some tips deposit Si, others deposit Sn, some alternate. Once a suitable tip is found, the process is reproducible. The authors have simulated the process using density functional methods and computing resources from the Barcelona Supercomputing Center, home of the beautiful MareNostrum machine (I couldn’t resist the temptation to include a picture).

MareNostrum,
Barcelona

The ‘Si’ structure took 1.5 hours to make. Most of this time was spent imaging, and some of this time was spent waiting for a new Si atom to appear on the tip. I say “appear”, because the tips recharge spontaneously, from material already on the tip, not by picking up atoms from the surface.

A fabrication process like this is far from being high-throughput manufacturing, and my bets are on self-assembly as a path forward, but the research illustrates several points:

• Mechanosynthesis isn’t restricted to processes like those found in ribosomes and nonribosomal peptide synthases: Physics allows mechanosynthetic processes quite different from those based on positional control of conventional reactive monomers.
• Fabrication processes are are often discovered, not designed. I doubt that the researchers expected to see this behavior, and the spontaneous recharging of tips is (to me) very surprising.
• For a wide range of applications, quantum chemistry works. Simulations showed how atom interchange can work, and simulations of the same kind could be used to discover and design other processes — which could seldom be implemented without a way to make atomically precise tips that meet the design specification.
• Techniques for atomically precise fabrication are accumulating. The first contributions came from chemical synthesis, more came from biotechnology, and yet others from materials science. Direct physical manipulation methods are becoming part of the toolkit.

Techniques for direct physical manipulation have thus far remained a laboratory curiosity, but they may gain practical value, and even now they illustrate direct positional control of the sort that will be characteristic of advanced mechanosynthetic methods.

See also:

• From Self-Assembly to Mechanosynthesis
• Molecular Assembly Lines
• High-Throughput Nanomanufacturing: Small Parts (with videos)
  

Delta Robot

I’ll get back to self-assembly and related topics soon, but at the moment, I’d like to show more about how macroscale manufacturing works today. There are strong analogies to engineering problems that will arise when a technology base is in place for building complex nanomachines, and I hope that even readers from manufacturing-oriented engineering cultures will enjoy thinking about familiar systems in this context.

In recent posts, I showed some high-throughput manufacturing processes, including videos of fast machines making small parts and of continuous-flow assembly machines that put small parts together with blurring speed. Products pour out fast, without a robot in sight. Somewhere on the spectrum between these fast, simple machines and slow, complex industrial robots are machines like the ones shown in the following video.

The first scenes show how parts are organized for assembly by blind, senseless machines: Feed machines extract parts from jumbled heaps, orient them, and send them on their way in single file. Each feed machine has only one key moving part: a bowl with cleverly shaped ramps and barriers — no machine vision systems, no robotic grippers. There are rough analogies between this and processes for organizing feedstock molecules for atomically precise nanomanufacturing.

The rest of the video shows how the parts get assembled:

In manufacturing, it makes sense to use fast, simple machines to make and assemble abundant, simple parts, then to use slower, more complex machines to put somewhat larger and more complex parts together, and, finally, to use fully programmable robotic mechanisms for short production runs and large, customized products. This principle is fully applicable to both macroscale and nanoscale processes; the numbers, sizes, and rates are different, but the pattern is the same.

Bonus videos: A fully programmable robotic positioner with an unusual geometry, high speed, and micron accuracy:

And here’s design based on similar principles, billed as the World’s Fastest Robot:

A device of this kind built at one-millionth this scale would execute moves at several megahertz. If you haven’t already seen it, you may enjoy the nanofactory video.