WO2003102756A2

WO2003102756A2 - Molecular-scale computational circuits based on charge-density-wave phenomena and methods related thereto

Info

Publication number: WO2003102756A2
Application number: PCT/US2003/016112
Authority: WO
Inventors: Edward A. Rietman; Bryan E. Koene
Original assignee: Triton Systems, Inc.
Priority date: 2002-05-28
Filing date: 2003-05-22
Publication date: 2003-12-11
Also published as: AU2003233638A1; AU2003233638A8; WO2003102756A3

Abstract

Discussed are computational circuits that include a matrix of computing elements, wherein the computing elements are at a length scale of nanometers. The computational circuits comprise charge-density-wave compounds for performing computations. Also disclosed are methods for reconfiguring the arrangement of mobile ions and electron density within a crystal lattice. The computational circuits and related methods of the present disclosure provide for significantly reduced circuit design size and significantly reduced manufacturing cost.

Description

MOLECULAR-SCALE COMPUTATIONAL CIRCUITS BASED ON CHARGE- DENSITY-WAVE PHENOMENA AND METHODS RELATED THERETO

Cross-Reference to Related Application

This application claims priority to U.S. Provisional Application No. 60/383,670 filed on May 28, 2002, the contents of which are incorporated herein by reference.

Field of the Invention

The present invention relates generally to the field of integrated circuits of electronic devices whose functional length scale is measured in nanometers and more particularly to molecular-scale computation circuits.

Background of the Invention

The state-of-the-art of molecular-scale computational circuit technology, as disclosed in U.S. Patent Nos. 4,580,110; 4,636,737; 6,128,214; 6,256,767; and 6,314,019, has some limitations. For example, no description to build working molecular-scale circuits exists in the art. Further, submicron circuits, the existing circuits that are the closest to being in the molecular-scale, typically require multiplixers and/or demultiplixers for sending data to the submicron circuit elements. These circuit elements are not at the nano-scale level, and consequently their "packing density" is below the packing density that a true molecular electronic circuit would have. The state-of-the-art submicron circuits typically require very short wavelength lithography for defining the placement of the conducting wires to which the molecular material will make contact. Furthermore, such integrated circuits are not rapidly reconfigurable or programmable.

Accordingly, there is a need for molecular-scale circuits that are rapidly reconfigurable, easily manufacturable, have truly nano-scale circuit elements, and do not require deep submicron lithography, or multiplixers and/or demultiplixers.

It is therefore an optional, non-exclusive object of the present invention to provide a computational system that comprises a matrix of computing elements, wherein the computing elements are at a length scale of nanometers.

It is also an optional, non-exclusive object of the present invention to provide a method for rapidly reconfiguring molecular electronic circuits that do not require deep submicron lithography in their manufacture. It is also an optional, non-exclusive object of the present invention to provide molecular electronic circuits do not require deep submicron lithography, or multiplixers and/or demultiplixers.

Summary of the Invention

The present disclosure relates to molecular-scale computation circuits. The present disclosure further relates to methods of building a molecular-scale neural network. It is generally recognized that one cannot build molecular electronics where the computational elements are individual molecules with molecular scale wires attached to them. However, the present disclosure provides new ways to exploit the physics of materials for computation, thereby circumventing these limitations. Specifically, disclosed herein is the use of low- dimensional semiconductors and mixed conductors (ionic and electronic) as computational substrates. The approach outlined herein significantly reduces circuit design and manufacturing costs, while reducing the size of the computational elements to the nanometer scale (i.e. molecular electronics).

In frequency- and amplitude-modulated AC electric fields, the ions in an intercalated compound can be moved around so as to achieve specific patterns. With fields about one volt ions can be moved to within 1 nm of each other. The ions can then be pinned with high frequency fields (about 10 MHz, for example). The clustered patterns of ions will cause charge density waves (CDW) and disrupt the lattice vibrations. These effects can be "measured" by high-frequency AC impedance. As noted below, Figure 1 shows an example of how a computational system based on charge density waves can actually be constructed. Certain predetermined electrodes would be used for data inputs and data outputs. Others would be used for programming the system with AC signals. The "programming" can be done with a genetic algorithm searching for a "circuit" that computes a predefined mapping relation. By measuring the input and output response and feeding back AC programming signals, a molecular-scale neural network can effectively be constructed.

Technology of the present invention is, in some respects, essentially a programmable crystal surface. The surface is programmed from wires (or bonding pads) attached to it. Other wires or bonding pads may act as inputs and outputs. When modulated AC signals are sent to the programming wires (or pads), mobile charge carriers are moved and can be patterned within the crystal lattice.

The molecular-scale computation circuits and related methods disclosed herein present a technologically-advanced and cost-effective process for integrated circuit fabrication. They are especially, although not exclusively, useful for vector-matrix multiplication and may be interfaced directly with molecular-scale sensors for signal processing. Other exemplary applications include target recognition, data compression, and process control.

Brief Description of the Drawings

Figure 1 shows a sketch of a thin film of a charge density wave material, K_xMoO₃ (0<x<l), deposited on SrTi0₃ to which gold electrodes have already been deposited. Films of charge density wave material will closely lattice match with SrTi0₃. The gold electrodes allow electrical contact to the charge density wave material.

Figure 2 illustrates effects of a) no field, b) a DC field, and c) a modulated AC field, respectively, on ions in a lattice consistent with the present invention.

Figure 3 is a schematicized view of a patterning of electrodes in connection with the system of the Example.

Figure 4 is a graph of impedance versus frequency measured in connection with the system of the Example.

Detailed Description

The computational system of the present invention includes techniques for exploiting the materials-physics of a class of substances known to exhibit a charge density wave. Charge density waves emanate from materials where electronic charge has been clustered by the external forces such as electric fields. Charge density waves typically are exhibited by materials that crystallize into layered structures, including but not limited to phosphates, phosphonates, bisphosphonates, sulfides, chalcogenides, metal oxides, oxy-chlorides, graphite, Tetracyanofulverene-Tetracyano-N-Quiniline (TTF-TCNQ), and layered double hydroxides. Often it is possible to intercalate ions between the molecular layers. These intercalated ions are mobile and can be manipulated by electric fields. With field strengths of about one volt it is possible to move the ions to within about one nanometer of each other. Frequency- and amplitude-modulated electric fields can be used to induce specific patterns of ions within the crystal lattice of the charge density wave material, thus effectively programming the Fermi level by inducing specific electronic modulations in the crystal lattice. It is to be understood that the electric fields serve to move electronic charge within the crystal with a resulting movement of cations, anions and/or holes. Small AC and DC fields that probe the Fermi level without significantly disturbing the electronic modulations in the crystal lattice affect the computations performed by the disclosed system. The computational system is targeted to special purpose computations involving vector-matrix multiplication such as pattern recognition, target prediction, and data compression, although it may be useful for other computations as well.

The use of low dimensional compounds (for example 2-dimensional layered structures) with ionic and electrical conductive properties provide interesting physical phenomena upon intercalation or insertion of mobile ions. Intercalation of ions within these compounds is a topotactic reaction, that is, the crystal structure does not change. This allows for a reversible modulation in the electronic structure of the material without a significant physical structural change. The reversible intercalation of a charged ion, for example, into a variable valence transition metal layered oxide framework results in a variation of the electron density of the compound. The position of the intercalated ion as well as the interaction with its closest neighboring ions will affect this charge density on a localized level. The movement of these ions within the structure to produce areas of higher and lower electron density within the lattice, can alter the Fermi surface of the material, and thus provide a method for programming a circuit on a molecular level.

Certain synthetic materials, known as band-gap materials, have been manufactured at the nano-scale by molecular beam epitaxy (MBE) techniques. The structures built at this scale force the Fermi levels to be at pre-specified energy values. For example, synthetic structures based on GaAs, AlGaAs, etc. will lattice match when prepared by molecular beam techniques. This technique enables construction of artificial materials with precisely located energy gaps in the Fermi surface. The same idea—generating engineered Fermi surfaces—is done with pure crystals of silicon, for example, by doping specific regions. At specific locations, small electrodes can be attached and electrons moved from one region to another by electrical signals that further change the energy level in specific areas. The end result, in the case of silicon devices, is transistors. With fully engineered and designed silicon structures computational circuits (e.g. CMOS logic) can be built. With the GaAs structures the same can be done, but because of the physics of GaAs, photonic devices for light emitting and detection at certain optical wavelengths may also be built.

In the solid state, molecules automatically "connect" to their nearest neighbor. This is also true for crystal lattices where the components may be atoms and ions instead of molecules. Above zero degrees Kelvin molecules (atoms and ions) are in constant motion. Defects in a lattice, or foreign molecular or atomic species, will disrupt the normal lattice vibrations. Similarly in amorphous materials defects and foreign species will affect the dynamic processes at that scale. But it is important to note that defects and foreign ions/atoms will not only disrupt the lattice vibrations but will also disrupt the Fermi surface.

If the crystal lattice is semiconductive and the defects and foreign species are ionic, then they will have significant impact on the Fermi surface and the lattice vibrations. In order to compensate for the fact that they are not in an equilibrium position the ions will create charge density waves in the lattice.

Certain materials, including but not limited to the insulating material molybdenum trioxide (Mo0₃), crystallize into sheets like layers that stack into three-dimensional arrangements. These layers are weakly held together and various ionic species, such as Li , Na⁺, K⁺, or Cs⁺ can be intercalated between the layers. These ions are weakly held between the molecular sheets and they can be moved in electric fields. This ionic movement is the basis of many solid state electrodes in rechargeable batteries.

Materials such as K_xMo0₃ (0<x<l) are excellent examples for these studies. Pure MoO₃ is white to yellow in color and is a large band gap insulator. The potassium-doped crystal (e.g. Ko.₃Mo0₃) is dark red to purple in color (with the darker color indicating an increase in electronic conductivity) and is a semiconductor, and the potassium ions disrupt the Fermi surface. In the absence of an electric field the ions will repel each other to a maximum distance and form a superlattice that may be commensurate with the underlying primary lattice. In the presence of electric fields the ions will move and the superlattice will now become incommensurate. The lattice vibrations in this incommensurate region of the crystal will send out quasi-periodic or even chaotic waves. These phenomena can be exploited for computation.

In particular, as shown in Figure 2, in an AC field the ions will drift back and forth at the frequency of the applied field. So in a frequency- and amplitude-modulated AC field the ions can be moved to within a nanometer of each other to form clusters and then be "pinned" into place with a higher frequency field (typically approximately 10-15 MHz and 1 Vpp). Naturally, these clusters of pinned ions would prefer to move away from each other and settle into some equilibrium state commensurate with the underlying molybdenum oxide lattice.

Without being bound by any particular theory, applicants believe each of these ionic centers forms a phase vortex ring capable of expanding until either a defect or another vortex ring is encountered. The rings may have diameter of up to hundreds of microns depending on the number-density of vortex rings. The rings likely behave like quasi-periodic oscillators known as sine-circle maps. A molecular-scale array of such rings would behave like a coupled map lattice—a type of computational architecture. The number-density, which can be controlled by the stoichiometry or by the electric field modulation, will essentially dictate the size of the rings.

These vortex rings may be "connected" with external fields and, at specific thresholds, may produce sliding charge density waves. Programming may be accomplished with a genetic algorithm searching for a "circuit" that computes a predefined mapping relation. This result is, arguably, analogous to a nano-scale programmable resistor/capacitor array or a programmable crystal lattice. By moving ions into discrete positions within the crystal lattice, differences in electronic conductivity for AC-based computational circuits may be exploited.

Further, the nanoelectric circuits of the present invention may be both electrically programmable and erasable. Hence, like certain existing programmable logic devices (e.g. Field Programmable Gate Arrays ("FPGAs")), the "circuits" of the invention may exist in software and disappear when the power is turned off and "reappear" when powered and reconfigured. If, therefore, a system were to be discovered by an unauthorized user, its circuits likely could not be reverse engineered. The circuits likewise would resist both tampering and differential power analysis, and any micro-code software could be encoded if desired to reduce the possibility of its being decoded by unauthorized personnel.

Example We have built a prototype system to demonstrate the ability to program the resistance values.

Gold electrodes were patterned (see Figure 3) on a small glass slide and a solution of the blue bronze (Ko.₃MoO₃) was deposited on the slide covering the electrodes and dried in a convection oven at 40C for 2 hours. Wires were attached; after that the device was encapsulated in epoxy resin to protect it from water vapor.

Electrodes 1 and 12 were connected to the positive terminal of a 15 volt source and electrodes 6 and 7 were connected to the ground terminal. The system was powered for 4 hours and then the AC impedance was measured across electrodes 3 and 4. Figure 4 shows the results before powering up with a DC field and after powering up with the DC field. At the lower frequency the impedance curves are flat. In these regions the impedance measurement was off scale. The graph indicates that at the higher frequency there is a significant and observable difference in the ac impedance. Between 100kHz and 1MHz the largest difference was observed. This is the basis of the programmable resistor array.

The foregoing is provided for purposes of illustrating, explaining, and describing exemplary embodiments and certain benefits of the present invention. Modifications and adaptations to the illustrated and described embodiments will be apparent to those skilled in the relevant art and may be made without departing from the scope or spirit of the invention.

Claims

What is claimed is:

1. A computational system comprising a matrix of computing elements, wherein the computing elements are at a length scale of nanometers.

2. The computational system of Claim 1, wherein the matrix of computing elements comprise modulations in the electron density of a crystal lattice.

3. A computational system comprising charge density wave compounds for performing computations.

4. The computational system of Claim 3, wherein the computational system is a vector-matrix multiplication system.

5. The computational system of Claim 3, wherein the charge density wave compounds comprise two or more layers and mobile charges.

6. The computational system of Claim 5, wherein the charge density wave compounds are intercalated.

7. The computational system of Claim 6, wherein the intercalated charge density wave compounds are selected from the group comprising phosphates, phosphonates, bisphosphonates, sulfides, chalcogenides, metal oxides, oxy- chlorides, graphite, Tetracyanofulverene-Tetracyano-N-Quiniline, and layered double hydroxides.

8. A method for reconfiguring the arrangement of mobile charges and electron density within a crystal lattice, comprising applying one or more external electric fields which cause an interaction between the mobile charges and the electron density to alter the Fermi surface of the crystal lattice.

9. The method of Claim 8, wherein the one or more external electric fields are changed to reconfigure the arrangement of mobile charges and electron modulations within the crystal lattice so as to perform calculations.

10. The method of Claim 9, wherein the one or more external electric fields are AC fields, and they are changed by modulating the amplitude and the frequency of the AC fields.

11. The method of Claim 10, wherein the reconfiguring is performed rapidly.

12. A computational system for performing computations utilizing the method of Claim 8.

13. A computational system for performing computations utilizing the method of Claim 11.

14. The computational system of Claim 3, wherein the computational system is for an application selected from the group consisting of programmable pattern recognition, target recognition, data compression, and process control.

15. The method of Claim 8, wherein the mobile charges are ions.

16. The computational system of Claim 5, wherein the layers are inorganic.

17. The computational system of Claim 5, wherein the mobile charges are selected from the group consisting of ions and holes.

18. The computational system of Claim 7, wherein the compounds comprise metal oxides.

19. The computational system of Claim 18, wherein the metal oxides comprise mixed-metal oxides.