The Center for Bits and Atoms
Activities and Findings
First Annual Report
CCR-0122419
June 15, 2002


Since the NSF's Center for Bits and Atoms was launched in the Fall of 2001 at MIT it has grown to involve 20 research groups across campus, with the work coalescing around four broad overlapping themes: computational mechanisms, biological interfaces, personal fabrication, and emergent engineering. Each of these areas addresses fundamental questions about the relationship between the information content in a system and its physical representation, and each draws on a number of disciplines to access grand-challenge problems at the bit-atom boundary. This report summarizes the progress made by CBA researchers in each of these areas, as measured against the goals for the first year, and sets milestones for the second year's activities.

Computational Mechanisms

Physical systems offer resources for manipulating information that go far beyond the assumptions of a digital programming model; as conventional device scaling approaches fundamental physical limits, the dynamics and interactions of these intrinsic degrees of freedom represent not just an obstacle but an enormous opportunity for manipulating both classical and quantum information.

The most significant milestone for the first year was:
  • implementation of factoring
This major goal in the development of quantum computing was completed by a team led by Prof. Chuang. In reference [1] they showed that the nuclear spin evolution in this molecule (a perfluorobutadienyl iron complex):
   

under irradiation by this sequence of ~300 RF pulses:



realized this quantum circuit:



which implements Shor's factoring algorithm to show that 15 = 5*3, in a polynomial rather than exponential number of steps. This is the largest quantum computation performed to date, in both system size (7 qubits) and algorithmic complexity (including a quantum Fourier Transform, and modular exponentiation).

Along with this experimental work, an expected theoretical goal was:
  • a theoretical foundation for embedding digital signatures into quantum algorithms
which was given by Prof. Chuang in reference [2], providing quantum security for one of the most important cryptographic algorithms used to authenticate information. An unexpected theoretical result in the first year of CBA was the discovery (in a collaboration between the Chuang, Gershenfeld, and Lloyd groups) that unitary operators (such as the RF pulses shown above) can be used to control non-unitary dynamics (such as thermal relaxation) in open quantum systems. Early experimental trials showed that by using intra-cycle spin programming it is possible to not just preserve spin magnetization long beyond the conventional T1 relaxation limit:

 

but also control its evolution (here tracing the letters "ML"):

 
This result, currently being prepared for publication, is important for developing quantum information-processing devices, but even more significantly it shows how a computational language can be valuable in understanding physical processes.

Even with this ability, however, quantum computing will not scale without the ability to design non-unitary dynamics for error correction and readout, hence this year 1 goal:
  • evaluation of the ability of ultrafast shaped optical pulses to enhance via hyperfine coupling the nuclear spin polarization used to represent logical states in complex molecules
This is being studied in collaboration with the NSF's FOCUS Center at the University of Michigan, jointly developing an experiment for ultrafast optical pulse shaping in a high-field NMR magnet with adaptive feedback, supported by a molecular modeling and analysis effort. Preliminary experiments have shown selective enhancement of molecular populations, which will be tested for their nuclear spin signature this summer.

CBA research is also exploring the power of coherence in classical as well as quantum systems. This year one goal:
  • Development and analysis of materials that can serve as physical one-way functions
was achieved in the Gershenfeld group by using mesoscopic light scattering. That project showed both theoretically and experimentally that, in the mesoscopic limit, speckle patterns provide a computational hash of the scattering structures. These data:



show the distance between 2400 bit (dashed curve) and 4800 bit (solid curve) keys extracted from glass microspheres in optical epoxy, for like (white) and unlike (gray) tokens. Given the milliradian correlation scale of the key, this corresponds to a bit-wise maximum entropy hash with on the order of a terabit of addressable keys committed to a penny's worth of materials. The enormous compression from the token information to the key information provides a physical mechanism for challenge-response queries by untrusted readers over insecure channels; these results are currently submitted for publication.

A growing focus in CBA has been appreciation of the power of analog classical dynamics to dramatically improve the performance of digital logic. This was anticipated in the first year goal of:
  • analog implementation of logical GF(2) dynamics, with application to efficient coding
This has been accomplished by the Gershenfeld group, in an Analog Feedback Shift Register (AFSR) that produced a maximal-length spread-spectrum sequence:

 

More importantly, the effort to explain this result has led to the recognition that an AFSR can be understood as a special case of an analog probabilistic network that performs digital inference; this idea is currently being developed both theoretically and experimentally. Overlapping work in the Sarpeshkar group has taken advantage of analog degrees of freedom to develop a time-based A/D converter:



and an analog memory:


that both promise to reach sub-mW efficiencies, using building blocks such as a mixed-signal amplifier with good power supply rejection for use in hybrid analog/digital devices [3].

Looking ahead, the goals for the next year in the Computational Mechanisms area are:
  • Coupled NMR and optical detection of molecular populations under ultrafast optical pumping
  • Quantum logic with greater than spin-1/2 dipolar coupling
  • Extension of entanglement under relaxation by control with unitary operators
  • Analog circuit implementation of graphical probabilistic networks for dynamical coding and decoding

Biological Interfaces

Real-time bi-directional interfaces to molecular information are particularly significant in a biological context, for medical applications certainly, but far beyond that for the promise of programming cellular machinery to make and manipulate nanoscale structures. Biomolecular assembly is far from diffusion-limited aggregation; nonlinear logic is essential to the workings of the assembly process, and learning how to interact with that machinery will be essential to matching the sophistication of its products. CBA's year 1 goals in this area were:
  • Nanoscale measurements of local molecular charge states at solid-liquid interfaces, for molecular readouts
  • Radio Frequency control of protein function through attached nanoclusters, applied to rudimentary nanoassembly
The first of these was accomplished by Prof. Manalis [4]. His group demonstrated that a microfabricated field-effect sensor located at the terminus of a freestanding cantilever:

 

can detect surface potential changes resulting from the adsorption of charged molecules in an aqueous environment.  The charge sensitive region, defined by lightly doped silicon, is embedded within the heavily doped silicon cantilever.  Since both the electrical trace and sensitive region are passivated with thermally grown silicon dioxide, the entire cantilever can be immersed in buffer solutions and cleaned with strong acids without degrading the electrical response. To estimate the surface charge sensitivity, they monitored the formation of a positively charged layer of poly-L-lysine on silicon dioxide.  In a 1 Hz bandwidth, the surface potential resolution is ~ 50 uV and the estimated surface charge resolution is ~ 3000 e/um2.

And the second of these goals was accomplished by a collaboration between the Jacobson, Zhang, and Hamad-Schifferli groups. In a landmark paper [5], they showed that gold nanoclusters can be attached to proteins:

 

(in this case, covalently coupling to the S-peptide in Ribonuclease S) which then under RF irradiation reversibly switch their conformation:


and hence their enzymatic activity:


 
Along with research on molecular input and output, work is also progressing in Prof. Slocum's group on MEMS structures for micro-fluidic control, based on creating nanoscale gaps that can meter, measure, and manipulate molecules of interest:


Next year, the goal on the input side is:
  • Demonstrate single-base selectivity of electronic DNA detection, with integrated sample handling
and on the output side:
  • RF switching of molecular assembly
  • Modeling of protein induction heating by nanocrystals
If these can both be accomplished next year, a final (ambitious) goal is:
  • Demonstrate feedback control of programmed molecular assembly using real-time electronic detection
This is a grand-challenge aim of CBA, promising to unite two of its fundamental research results into a new field at the intersection of biology, chemistry, electrical engineering, and computer science.


Personal Fabrication

Another grand challenge problem for CBA is to bring the malleability of digital worlds to the physical world, through table-top means to fabricate active logic along with sensors, actuators, displays, and three-dimensional mechanical structures. Just as the packaging of Personal Computers provided the capabilities of mainframes for ordinary people, with revolutionary consequences, this effort aims to create Personal Fabricators that make accessible the capabilities of industrial electronic and mechanical fabrication technologies.

This project is based on fundamental materials development, including chemistries for printing wires and switches. The earlier discovery by Prof. Jacobson's group of "nanotectic" chemistries for printing inorganic semiconductors inspired these year 1 CBA goals:
  • all-printed logic circuits using nanotectic chemistry, starting with ring oscillators
  • active three-dimensional electronic structures, fabricated by additive writing processes
  • three-dimensional design tools to support direct printing and writing of electronic materials
These activities are described in references [6] and [7]. Ink-jet printing and liquid embossing processes were developed for use with a range of nanoparticulate systems, which were used to print structures including transistors, three-dimensional vias, coils, motors, and actuators in a table-top process compatible with plastic substrates:



(showing an actuator on the left, and motor on the right). These represent the first use of all-printed technologies to make MEMS devices, and offer a straightforward route to integrate a range of active materials. One promising candidate being developed by the Swager group in the last year is based on rotaxanes, which are structures held together by steric constraints rather than covalent bonding. In this polymer:


the large tetraphenylmethyl groups cannot squeeze through the macrocycles in which they are trapped, thereby self-assembling a macroscopic network of insulated molecular wires [8]. The metal binding site can then be used for electroactive switching, as well as other transduction mechanisms.

Along with development of materials and printing mechanisms, CBA is working on the design tools to support personal fabrication. The Mikhak group has developed "LaserLogo," an extension of the Logo programming language designed to operate table-top fabrication machines (such as those being deployed in the "fab labs" described in the Outreach section below):



This not only provides an intuitive visual design interface, usable by non-traditional users such as young children, it is based around the kind of programmable parametric representation that is currently found only in high-end CAD tools. This capability is essential for being able to express the complex functional relationships in materials and mechanisms in a Personal Fabricator.

In the next year, the Personal Fabrication effort has a goal of integrating the enabling components that have been developed to date:
  • Demonstrating with table-top fabrication technologies of the integration of input, logic, and output devices
This will be supported by use of the internally-developed CAD/CAM/CAE software:
  • Control of both research and field fabricators by user-programmable design tools
along with consideration by a new CBA researcher, Prof. Sass, of personal fabrication design issues over much larger length scales.

Emergent Engineering

The preceding projects for printing, growing, and synthesizing information-processing devices together promise to make it economically feasible to scale systems from millions to billions to trillions of components. The only thing that is certain about this evolution is that current technological design practice will fail for systems of such enormous complexity. In domains ranging from chip design to network design, managing complexity is itself rapidly becoming one of the most severe scaling constraints, with countless examples of emergent failure mechanisms, but few insights into emergent success mechanisms. The last broad research area for CBA is considering how to design principles by which complex systems come to function, without explicitly specifying or understanding the details of how they actually work. This effort seeks to bring the rigor with which attributes such as power, bandwidth, and signal-to-noise are currently understood to properties such as emergence, adaptation, and hierarchy.

The first embodiment being investigated in some depth is "paintable" computing, which was the subject of two theses in the last year [9, 10]. This project seeks to program enormous numbers of randomly connected imperfect components, rather than a single enormous perfect component (as assumed by current chip fab practice). The aim is creating computing that is literally fungible, so that it could be added by the pound or square foot to improve performance as needed. The key insight making this possible has been the development of a probabilistic local shared-memory programming model for mobile code. One of the year 1 goals was a
  • device-level simulator
This has been developed by Dr. Butera and is now being packaged for wider distribution. In this example of it:



each dot is running a model of one of the paint particles, with short-range communication to its neighbors. The example shows a computational front propagating through the medium that is configuring it for point-to-point wiring in an ad-hoc network. Unlike prior work on ad-hoc networks, that capability is emerging here as just one application on a much more general-purpose computation and communications substrate.

Along with the modeling effort, a second year 1 goal was:
  • a functional simulator for a "paintable" computer, along with exemplary programming implementations, adequate for guiding device design
This was accomplished with a "pushpin" model, developed by the Paradiso group. A small stack of boards serves as a functional emulator of the paint particles using short-range optical communications, with pins extending into a conducting substrate providing power:

 
 
Demonstration applications from Dr. Butera's simulator are now being ported to the pushpin platform. It is also being used as a testbed for this year 1 goal:
  • distributed time synchronization developed as a "delayering" test case, exposing physical-layer channel resources to the application layer, with a focus on sensor network requirements
as described in reference [11].

An important goal for CBA is to build emergent engineering properties into systems that preserve compatibility with the existing Internet infrastructure; this was reflected in year 1 goals:
  • delayer distributed command-and-control algorithms
  • scalable ultra-lightweight implementations of Internet Protocol and its physical transport, for provably-correct serverless command-and-control of building infrastructure, used as a test-bed for language abstraction for programming emergent systems
These ideas were explored through a set of components that were developed in the Gershenfeld group (with Dr. Shrikumar) and used in testbed installations, and are eventually destined for the new labs that will house CBA [12]:



Each device provides a complete set of Internet protocols (ARP, ICMP, IP, UDP, TCP, HTTPD), in a few dollars of parts and a few kB of code, by implementing their phenomenological functionality without the overhead of inter-layer abstractions. Each device also contains the data records associated with it, along with the algorithm threads needed for updating them, so that assembling the devices simultaneously builds a network, a distributed data structure, and a parallel computer to manipulate it. This first embodiment used the RS485 wiring already used for building control systems to carry these native IP messages, in a multi-drop network that did not require impedance matching. These components, along with other wired and wireless transports, are growing into an emerging standard for low-data rate IP, working closely with a network of corporate partners including Sun, HP, Microsoft, Intel, Motorola, and United Technologies.

Having abandoned strict layering as an organizing principle for such distributed systems, it will nevertheless be essential to partition functionality to obtain usable scalability. Dr. Sollins is using these ultra-lightweight networks as a domain for studying that question, in the context of the larger NewArch project. She is introducing an abstraction (called "Regions) of scoping mechanisms, with functional modules replacing layers to organize capabilities. In the coming year these ideas will be developed in the context of routing. Dr. Sollins has also been considering questions of privacy and security for such ubiquitous connectivity.

Finally, complex adaptation is being explored in the context of robotics by Profs. Breazeal and Seung. The Breazeal group is developing sensor "skins" with local low-level information processing, along with novel transmissions/actuators, and the Seung group has developed physical models that will be used for reinforcement learning of mechanical control systems [13].

The goals for the next year in Emergent Engineering are:
  • Chip-level design of a paintable computer, and mutli-chip-module level emulation
  • Introduction of distributed security and routing into the embedded IP infrastructure, and analysis of system scaling
  • Adaptive learning of mechanical controls using distributed sensing and actuation

Outreach
 
The CBA outreach effort looks beyond teaching people about science, to focus on enabling them to do science, for both its practical and intellectual impact. The overall year 1 goal was:
  • domestic and international proof-of-principle tests
This has been accomplished with a network of partner organizations, including the Computer Clubhouses in under-served neighborhoods, the PIE network of museums, and the Media Lab Asia initiative based in India that is developing and deploying information technologies for global development. This goals had 5 sub-themes:
  1. embedded processors with an accessible programming model for sensor interfaces, for instructional and environmental measurements
The initial devices developed for this effort in the Gershenfeld group include a UV-VIS spectrometer, and RF complex materials impedance analyzer, shown here with one of the rural labs where they will be tested:



These use low-cost consumer electronics components, and are aimed at application in rural environmental, healthcare, and food labs. The ~$100 devices have been calibrated in the lab against ~$100k instruments, and are now beginning field testing.
  1. low-cost platforms for computation, communications, instrumentation, and fabrication for local technological design and production
This project, led by Dr. Mikhak, is developing the "Towers" system of stacking circuit boards with microcontrollers, network interfaces, displays, sensors, and actuators, supported by visual dataflow programming tools. The aim is to provide enough functionality in the kit to be able to assemble a wide range of computing and communicating capabilities, including the means to make more system components. These boards have been fabricated and debugged, and production for deployment is now ramping up.
  1. Printed Circuit Board layout and mechanical design tools to make rapid prototyping available to non-technical users
These will be provided in "fab labs" that CBA is setting up under the supervision of Dr. Mikhak, using simple optimized NC fabrication technologies (such as a computer-controlled knife used to trace out circuit boards), initially in five sites that include both inner-city and rural domestic and international settings. The rapid prototyping tools are being provided with the Towers and the appropriate analytical instrumentation, to bring technological design and experimentation out of the laboratory and to non-traditional users. This field deployment is complemented by a program led by Prof. Slocum to bring in students from under-served communities to use CBA resources.
  1. field trials in underserved communities in both developed and developing countries  
Initial experiments were performed in the network of Computer Clubhouses, supervised by Prof. Resnick [14]. For example, in one of these sites in a slum in Delhi, a computer-interfaced microscope was provided:



The boy shown was horrified when he saw what was in the local water supply, and began experimenting with boiling drinking water at home:



to remove pathogens. He is now working with his friends on a movement to educate their families about the importance of proper water treatment.
  1. collaborative tools for distributed problem-solving in technology for development
This work is being supervised by Prof. Pentland, who is leading the Media Lab Asia project. That includes a Geographical Information System for village applications, a medical data system for use by rural nurse midwives, and very low cost store-and-forward data transmission using physical transport such as postal vehicles. These large-scale field activities are complimented by a student-developed and -run Web site at MIT [15] for "open-source" technological problem solving by bringing global development challenges to science and engineering students around the world (including those in CBA).

For more advanced students, Profs. Lloyd and Chuang have been working on making quantum information (and quantum mechanics) more accessible. Prof. Lloyd has developed interactive role-playing games, tested in local schools, to teach the workings of a quantum computer, and Prof. Chuang has developed a pioneering unit for MIT's Junior Physics Lab that lets the students actually perform quantum computations.

These thrusts will be continued in the Outreach goals for the coming year:
  • Equipping field "fab labs," and developing pedagogical models for their use in local technology development
  • Deployment and testing of applications for appropriate information technologies and analytical instrumentation in global development

 
Operations

A major component of CBA is the infrastructural investment:
  • Initial investments in shared experimental infrastructure, with a focus on nanoscale assembly and instrumentation development
Under the supervision of Prof. Jacobson, the major acquisition in the first year has been a combination of a dual-beam Focused Ion Bean Writer and SEM, along with an Environmental SEM with materials analysis, to create a unique nanostructure rapid-prototyping suite:



These tools together are typically found only in industrial settings where they're needed for mission-critical applications such as mask repair; CBA will provide the first facility where they are available for non-traditional users, in the spirit of its work on rapid-prototyping on longer length scales. In the coming year, attention will turn to improving access to the fabrication of silicon micro- and nano-structures, working with Profs. Schmidt, Manalis, Jacobson, and Slocum on developing capabilities including mask making, anisotropic etching, and confocal imaging.

The growth of the technical infrastructure was matched by the growth of the intellectual infrastructure, including the goal of:
  • Inaugurating a visitor and lecture series
which was based around a weekly speaker series, and:
  • Ongoing operational oversight by a Technical Advisory Board
with the initial meeting of this group including:

   Noubar Afeyan (http://www.newcogen.com/wwa_mt.asp#Afeyan)
   Ruzena Bajcsy (http://www.berkeley.edu/news/media/releases/2001/10/04_Bajcsy.html)
   David Dalrymple (http://www.umbc.edu/window/dalrymple.html)
   John Doyle (http://www.cds.caltech.edu/~doyle/home.htm)
   Barrie Gilbert (http://www.edtn.com/analog/barrie2.htm)
   Alan Huang (http://www.jp.apan.net/IWS2000/alan-bio.html)
   Peter Shor (http://www.research.att.com/~shor/)
   Susan Watson (http://www.middlebury.edu/~physics/meet.html)

Over the coming year, the external advisors are expected to spend time at CBA for longer-term visits. Finally, the goal of:
  • World Wide Web presence with external interfaces into the program
will be met by developing http://cba.mit.edu from a simple event log into a research tracking and presentation resource.

References
  1. L.M.K. Vandersypen, M. Steffen, G. Breyta, C.S. Yannoni, M.H. Sherwood, and I.L. Chuang, "Experimental Realization of Shor's Quantum Factoring Algorithm using Nuclear Magnetic Resonance," Nature (414), pp. 883-887 (2001)
  2. D. Gottesman and I. Chuang, "Quantum Digital Signatures," http://arxiv.org/abs/quant-ph/0105032 (2001)
  3. M. O’Halloran and R. Sarpeshkar, "A Low Open-Loop Gain, High-PSRR, Micropower CMOS Amplifier for Mixed-Signal Applications," Proc. of the IEEE Conf. On Circuits and Systems, (Vol. II), pp. 424-427 (2002)
  4. E.B. Cooper, J. Fritz, G. Wiegand, P. Wagner, and S.R. Manalis, "Robust Microfabricated Field-Effect Sensor for Monitoring Molecular Adsorption in Liquids," Applied Physics Letters (79), pp. 3875-3877 (2001)
  5. K. Hamad-Schifferli, J.J. Schwartz, A.T. Santos, S.G. Zhang, and J.M. Jacobson, "Remote Electronic Control of DNA Hybridization Through Inductive Coupling to an Attached Metal Nanocrystal Antenna,"  Nature (415), pp. 152-155, (2002)
  6. S.B. Fuller, E.J. Wilhelm, and J.M. Jacobson, "Ink-jet Printed Nanoparticle Microelectromechanical Systems," Journal of Microelectromechanical Systems (11), pp. 54-60 (2002)
  7. C. Bulthaup, E. Wilhelm, B. Hubert, B. Ridley, and J. Jacobson, "All-Additive Fabrication of Inorganic Logic Elements by Liquid Embossing," Applied Physics Letters (79), pp. 1525-1527 (2001)
  8. M.J. MacLachlan, A. Rose, and T.M. Swager, "A Rotaxane Exciplex,” J. Am. Chem. Soc. (123), pp. 9180-9181 (2001)
  9. J. Lifton, "Pushpin Computing: a Platform for Distributed Sensor
    Networks," M.S. thesis, MIT (2002)
  10. W. Butera, "Programming a Paintable Computer," Ph.D. thesis, MIT (2002)
  11. J. Lifton, D. Seetharam, M. Broxton, J. Paradiso, "Pushpin Computing
    System Overview: a Platform for Distributed, Embedded, Ubiquitous Sensor
    Networks," Proc. of the Int. Conf. on Pervasive Computing, Zurich (2002)
  12. http://www.media.mit.edu/physics/projects/IP/bldg/bi/
  13. R. Tedrake and H. Sebastian Seung, "Improved Dynamic Stability using Reinforcement Learning," 5th International Conference on Climbing and Walking Robots (CLAWAR), Professional Engineering Publishing Limited (2002)
  14. M. Resnick, "Rethinking Learning in the Digital Age," in "The Global Information Technology Report: Readiness for the Networked World," edited by G. Kirkman, Oxford University Press (2002)
  15. http://thinkcycle.org